Computer Networks and Inventive Communication Technologies

The study develops the mapping-based cuttlefish optimization algorithm (MCFA) to select the cluster head (CH). A local and global search integration is occurred by using this algorithm, and chaos mapping is introduced to solve the issues of trapping in local optimum. NS-2 tools are used and test the efficiency of the MCFA algorithm with ant colony optimization (ACO) in terms of path optimality, delay time, etc. The results prove that the proposed MCFA achieves better performance than the existing ACO algorithm.

Due to Internet, vast amount of data is generated day by day; from those data to find useful insights, there is need to identify and extract the subjective information. Today’s trends show that people are buying any products or watching any movie on Web sites, and they write the feedbacks related to that product or movie, which will be helpful to business in terms of profit. For that, the need is analysis of written reviews which will be done by sentiment analysis. It is a method which is used to gauge opinions of individuals or groups of persons related to their products or movies. This method will extract the meaningful insights from the written reviews in the form of positive, negative, or neutral. Analysis of sentiment is also known as opinion mining. In this paper, hybrid deep learning model (CNN + LSTM) is applied on IMDB movie review dataset and performs a comparison with CNN model.

From closedown of December 2019, coronavirus has directly exhibited a lofty rate of transmission, coercing the World Health Organization to contend in the month of March 2020 that this unbeknownst coronavirus can be depicted as a pandemic. COVID-19 epidemic has guided to an operatic misplacement of deathly life over the public and presents an unbeknownst complaint to public fitness. It also affects the food systems of the person and the world of work. Once the person is infected by COVID, the metabolic exertion of vulnerable cells in his or her body is enhanced, similar as the one driven by COVID-19. The country’s dietary habits are analyzed to predict the particular person’s death rate. By using KNN algorithm, the performance metrics such as accuracy, precision, recall, and F1 score are evaluated for the country’s dietary habits. In this research, both clustering and classification are combined to increase the accuracy of the prediction of death rate of the person. K-means is used for the clustering of the countries, and KNN is used for classifying the countries. The 170 countries are clustered based on the country’s dietary habits, and other disease affected rate using K-means clustering algorithm. Countries are clustered into high and normal death rate countries based on the country’s dietary habits and another cluster into high and normal death rate based on the other disease affected rate rather than COVID-19. Using the country’s dietary habits and other disease affected clusters, the death rate of the person is predicted. After clustering the data based on the country’s dietary habits and other disease affected rate, the KNN algorithm is used to classify and identify the person’s death rate. Using clustering and classification algorithms in a combined way, an accuracy of 79% is achieved.

Modern communication networks have grown highly complex and dynamic, making them difficult to describe, forecast, and govern. So, the software-defined networks (SDNs) have emerged. It is a centralized network, and it is flexible to route network flows. Traffic engineering (TE) technologies are used with deep reinforcement learning (RL) in SDN to make networks more agile. Different strategies for network balance, improvement, and minimizing maximum link usage in the overall network were considered. In this article, recent work on routing as well as TE in SDN and hybrid SDN is analyzed. The mathematical model and algorithm used in each method are interpreted, and an in-depth analysis has been done.

With an ever-increasing number of user equipment (UEs) and rising bandwidth demands of new applications, deployment dense heterogeneous cellular networks have been embraced in various network scenarios. The cells experience unloaded due to the random UEs mobility and cells deployment, which degrades the network performance such as handover success, throughput, and load distribution. We propose an enhanced cluster-based algorithm for small cell mobility load-balancing networks to address such a problem. The conventional mobility load-balancing (MLB) algorithms consider only the contiguous neighboring cells and do not expand enough performance of the network, while other MLB algorithms consider the neighboring cells of the total network experienced unneeded MLB actions. The proposed load-balancing algorithm studies overloaded cells and neighbors using the proposed efficiency parameter. To begin with and to identify the overloaded cells, we propose an efficiency factor B that compares a pre-defined threshold and the network threshold to control the algorithm triggering by modifying the CIO parameters of the cluster cells in both medium-loaded and overloaded cells triggering in both medium-loaded and overloaded cells. Then, to control the distribution, we propose a method to shift only a portion of the serving cells load, so the target cell load after handover always be equal to or less than B. In a low UE speed scenario, the simulation results showed a lower standard deviation (SD) by 19.99% and enhanced throughput by 7.355%.

Irony is a sarcastic term that is used to actively criticize people. Irony is frequently expressed through indirect phrases. Politics, sports, business, and social media are the different places where ironic statements can be found. Some people can perceive irony in words, while others are clueless about how others have expressed themselves. To recognize these elements of speech, various machine learning and deep learning algorithms are applied. The main purpose of this study is to propose a broad summary of the ironic and non-ironic words found in subreddit data were used to categorize the text comments. This manuscript describes about the well-known datasets that are used to create the ironic detection networks by providing a major focus on classifying the comments. The considered dataset is partitioned to make the task easier. Finally, the process used for identifying irony phrases and the most successful classifiers and evaluation criteria are also discussed.

IoT sensor/edge nodes are vulnerable to many well-known threats. Maintaining the absolute reliability of any IoT LAN requires validating edge nodes before entering a network, mainly post entering a sleep state. These IoT nodes possess limited resources and measurable limits, making this a complicated issue. IoT equipment is frequently exposed to the elements since many IoT installations occur in unregulated settings. Equipment cloning and the theft of private keys from edge nodes are among the most common threats on IoT networks because of their easy physical access. These issues motivate to design a highly authentic device that intends to connect with other devices to preserve the stored and incoming data. When the authentication information matches, the connection is established with requesting device. This work aims to create an ultra-thin verification mechanism for IoT LAN-connected end equipment. The gateway, which serves as edge computing equipment, authenticates the end-user equipment. Formal and informal safety checks are performed on the recommended verification process. The simulation is done in MATLAB 2020a environment, where the proposed delicate authentication scheme works well with superior outcomes.

The detection of plant disease is an important problem that requires concentrating on the effective production of agriculture and the economy. The traditional techniques were used to detect this plant disease, but it is a difficult task that needs a lot of time, work, and expertise. The recent research gathered more attention recently between researchers, practitioners, and academicians called automatic detection of plant disease. This automation uses two techniques called machine learning and deep learning that helps for the identification of plant disease at the earlier stage as it finds in leaves of plants. A complete examination was done for the evaluation using the modern study on the possibility of acquiring the machine learning models to find the plant disease. There are four methods of diseases and infection on crops. Primarily, various possible diseases and infections on various crop types are investigated with the cause for the happening and feasible symptoms for the identifications. A thorough investigation on the various pace is needed that is evolved in the detection of plant disease and the categorization with the help of deep learning and machine learning that are given. Different datasets that are there in online to detect plant disease are also provided. A complete investigation regarding different machine learning and deep learning depending on the classification models is discussed, which are given previously and are suggested over the world by various researches for the four above-mentioned crops related to the evaluation on performance, the used dataset, and the method of feature extraction. Finally, different difficulties are listed and presented when using machine learning and deep learning techniques to identify the plant disease and present the future research scope.

The FinFET generates a lot of effort, which leads to a higher level of control in the conditional channel. Despite its low performance, 6T static process of random-access memory is modified by circuit function design. This increases the speed of static random-access memory by reducing the bit line with loading effect. With a standard level of cell, 6T static random-access memory face challenges in terms of stability and degradation. It causes disruption in low-level power mode. With a reduced low level of threshold voltage, 6T SRAM has difficulties with output level voltage. In 6 T-type SRAM cell, the conditional transistor destroys the entire read operation. In 6 T SRAM cell, noises destroy the stored level data in nodes, which establishes a direct path between the bit line and storage nodes. We must overcome the 8T-level SRAM as a cell in their recommended read-level stability. To improve the 8T-level SRAM read level of stability, the FinFET with 6T and 8T for conditional achievement of SRAM in 8T level for static random-access memory, and 10T type for SRAM with cells has been increased to improve the need to compare different results by using micro wind as a tool.

Intelligent reflecting surface (IRS) has been materialized as a new physical layer technology to reconfigure and intelligently control the wireless propagation environment. However, for improved system performance in the presence of IRS, knowledge of proper channel state information (CSI) is essential. In this paper, we present a novel and efficient channel estimation technique based on sparse Kalman filter (SKF) in intelligent reflecting surface-supported (IRS) multiple input multiple output (MIMO) millimeter wave (mmWave) wireless communication systems. In addition, we mitigate the training overhead by taking advantage of the sparse attributes and low-rank behavior of the IRS-mmWave MIMO-cascaded channel. Channel effect is evaluated with Rayleigh fading and QPSK as the modulation technique. Simulation results demonstrate that our proposed sparse Kalman filter-based channel estimation method outperforms the conventional methods in the matter of channel estimation accuracy, bit error rate performance, and reduced training overhead.

Multipliers are widely used in the arithmetic units of digital signal processors. The objective is to design Bit Pair Recoding technique using M-GDI, CMOS technology and to analyze the performance of Bit Pair Recoding technique in terms of area, power, and latency. The methodology of the project consists of a Bit Pair Recoding technique as a top module. In the first step, the pre-encoder is designed for Bit Pair Recoding. The second and third steps involve developing an encoder to generate encoded signals and developing a decoder that generates the partial products for the respective encoded signals, a multiplier register to hold the data from the data bus, and developing a carry propagation adder for the purpose of adding partial products. The Bit Pair Recoding technique is designed using CMOS 45 nm technology and M-GDI 45 nm technology. The power consumption of the design with M-GDI 45 nm technology using a pre-encoded mechanism is about 6.0197 mW. The Bit Pair Recoding technique with M-GDI 45 nm technology provides 17.82.

Object detection is the technique of locating targets in an image scene. Moving object detection is to track the object in successive frames which forms the video. Moving object detection forms the primary step in surveillance applications and traffic monitoring systems. With the advent of deep learning remarkable performance is observed in object detection. But the following challenges of detecting small and low-resolution moving objects in the region of interest (ROI), objects with infrequent motion still remain. Deep learning background detection techniques aim at cancelling global camera motion and background movements. The review focuses on the strengths and weaknesses of various deep learning-based moving target region proposal networks and background separation methods.

Dengue is a mosquito-borne, deadly viral disease that is a major threat to public health all over the world. Dengue and covid-19 symptoms are almost same, and sometimes, people are confused about which disease they are infected with. This year in Bangladesh dengue and covid-19 patients have been increasing at an alarming rate, and most of the time people didn’t properly recognize the disease. A developing country like Bangladesh has faced many difficulties to handle this situation. The target of this research work is to analyze the symptoms and predict the chances to get infected with dengue fever. Machine learning techniques are widely utilized in the health industry to detect fraud in treatment at lower cost, predictive analysis, cure the disease. Four machine learning algorithms are used which are support vector machine, decision tree, K-nearest neighbor, random forest to predict dengue fever based on symptoms. The results were compared for percentage split and K-fold cross-validation method for before and after applying principal component analysis. The experimental result shows that the support vector machine algorithm provides the highest performance compared to others algorithms.

Online shopping becomes indispensable among the people worldwide. Clickstream, collaborative filtering and machine learning algorithms play a considerable role to analyze the browsing behavior and predict the next click of the customers. In this research, k-nearest neighbor is applied to classify the customers into three groups: Regular, Special and Exceptional. User-Item-Detail matrix is constructed to identify the similarity among the online customers. Exclusive recommendation is provided to the customers based on user classification. The accuracy of the research is evaluated with the parameters precision, recall, and f-scores.

The Android operating system has changed what mobile can do. The population of Android mobile users has been increasing, and the number continues to grow as the number of mobile users increases. Primarily, the applications used in Android devices are distributed via the Google Play Store. These applications are required to meet several criteria in order to be distributed via the Play Store. However, attackers sometimes find their way to compromise the Android operating system devices and steal users’ sensitive information. Malicious software or malware is designed to extract data from users by compromising their devices for financial gain or other reasons. This study uses machine learning algorithms to detect Android malware in the Android malware dataset. Using CICAndMal2017 dataset, random sampling was done to extract a balanced dataset. Feature engineering was performed to obtain the most significant features. Machine learning algorithms were trained using the balanced dataset with selected features. Initially, all the models were trained using ‘Label’ as the target variable and then ‘Family’ as the target variable. Maximum accuracy of approximately 99% was obtained using random forest in both cases.

Fundus images are mainly used for the screening and detection of retinal diseases. Defect in the eye’s posterioir blood vessel causes Diabetic retinopathy. Proliferate Diabetes Retinopathy and Proliferate Diabetes Retinopathy are the two stages of diabetes that develops over the time. The retinal fundus image is commonly used in clinics to diagnose and treat a variety of eye disorders. It is also one of the most important resource for diabetic retinopathy mass screening. In this paper, a survey has been made on various techniques available in image processing such as morphological operations and texture analysis to identify diabetic retinopathy. This survey extended to investigate machine learning and deep learning approach to extract various features namely, area of hard exudates, microaneurysms and blood vessels to identify and detect the depth of the disease. Finally, the review has provided a summary and analysis of most of the recent techniques adapted to detect diabetic retinopathy.

In this paper, a Delta Lake workspace is created by using Azure portal; whereas, the ADLS Gen2 (ADLSG2) acts as primary storage account with a container to store workspace data. Despite the hype, ADLSG2 is immutable and it cannot perform analytics. This drawback leads to the introduction of Azure Databricks Delta Lake (ADDL) to fascinate the learning pattern that can be utilized for developing a support system to anlyze stock market and initiate better prediction on forecasted stock price. Databricks enhancement is an open source named Delta Lake, which remains as a pipeline for atomicity consistency isolation durability (ACID) table storage layer over cloud object stores. Finally, the ADDL performance is evaluated with the existing big data platform by using the parameters like memory usage and CPU usage.

Because artificial illumination is created in an indoor environment, multimedia data is recorded in an optimal atmosphere. On the other hand, in an outdoor setting, it is critical to eliminate the weather influence. Outdoor vision devices are used to capture scenes in surveillance. Due to the enormous size of the drips, the items will get motion blurred in changing weather conditions. Several computer vision techniques that employ feature information, such as object identification, tracking, segmentation, and recognition, will be harmed by these disturbances. Even if only a little portion of the object is obscured, the object cannot be accurately tracked. Rainfall features that pixel image is not entirely surrounded by rain in all data. The dynamic adverse weather model is researched for restoration resolution. Rainfall is a very important part of a bad weather system. Rain-fed energy has a strong local structure and is strongly influenced by backlight. When light passes, it is repeated and visible, shining in the surrounding area. The movement fades, when it falls at a high rate. The vehemence of the rain line is therefore, determined by the light of the descent, the radiation of the backdrop, and the unification time of the camera. Particles of rain and ice are very difficult to analyze. Rain-like spatial and temporal occurrences can be produced by some scene dynamics. In this work, we will look into ambient light estimation for submitted multimedia material using image processing algorithms. With in-depth reading techniques, we improved light measurement and used the histogram measurement method.

During the recent COVID-19 outbreak, many people opted to stay at home and completely run their businesses online. The recent research studies show that, the frozen food sales and social media usage have grown exponentially as a result of the change in people’s behavior and companies’ marketing strategies. Frozen food sales are expected to grow as a result of social media marketing, and also, the relationship between buyers and sellers is becoming more flexible. The main purpose of this research study is to examine the recent increase in frozen food sales as a result of social media’s influence. The proposed research work has been carried out by using qualitative technique, and also, this research study has conducted a literature analysis to learn from previous research works based on what factors they contribute to increase sales, and why social media is the best medium for marketing. The literature study suggests that frozen food sales rise as a result of consumer behavior, and that social media is the best platform for marketing their products. The results indicate that the behavioral changes have a significant impact on frozen food sales, and that social media assists in product marketing and enhancing the key drivers of growing consumer interest. As a result, a large number of frozen food companies are considering the option of utilizing social media as a marketing tool.

The recent advancements in computer vision-based innovations like image identification and object detection are attracting the attention of the modern research community present across the globe. Data analysis domain is interesting, particularly when it comes to image augmentation and categorization. Image recognition will operate on a variety of variables in different contexts relating to human behaviour, such as what sort of behaviour someone would display when driving a car. Activation such as network expanders is used in numerous significant neural network models, which are used in image processing domain. Learning the intricacies of features is gradually becoming more prevalent due to the fact that nearly all mobile devices are connected to the Internet and to the fact that virtually all portable devices are connected to the Internet. To achieve this, the visitor obtains a digital camera and uses it to capture the marker and upload the images to a nearby server where they may be compared to the photo archive. Some useful information can be provided during an image search. For example, learning more about significant landmarks, local trails, or nearby areas of interest, or interesting features are also given on the match to be searched. The broader and deeper form of application will be better for guiding people around the region. This paper proposes an enhanced convolution neural and deep convolution neural network-based method for identifying different landmarks by using machine learning [ML] algorithms. The system uses spatial extreme learning machine (SELM) for performing the optimistic feature recognition and enhancement.

The Housing Society Management System aims at managing the day-to-day activities of a cooperative society and making the current situation in society simple and efficient. The proposed research focuses on creating online apps for handling complaints in gated and guarded communities. These community areas have fencing and are supervised by a guard. They are managed by a company that is paid to look after the assets, amenities, and security in residential communities. Text messages and WhatsApp are now the only complaint handling methods available. As a result, it is difficult for management to review the complaints, and the entire process takes longer. In addition, residents are unaware of the actual status of their complaints. The suggested system entails the application of web-based applications to manage this. The data sent is stored in the developing application’s integrated database. This application is put to the test on-site to see how effective it is. The findings of the poll revealed that the application is successful in dealing with resident problems.

In daily life, we come across tons and tons of information which can be related to news articles or any kind of social media posts or customer reviews related to product. It is difficult to read all the content due to time constraint. Being able to develop the software that can identify and automatically extract the important information. There are two types of summarization methods. Extractive text summarization is the method where it picks the important content from the source text and gives same in the form of short summary, and on the other hand, abstractive summarization is the technique where it gets the context of the source text, and based on that context, it regenerates small and crisps summary. In this paper, we use the concept of neural network with attention layer to deal with abstractive text summarization that generates short summary of a long piece of text using review dataset.

A customer is a churner when a customer moves from one service provider to another. Nowadays, with an increasing number of severe competition with inside the market, essential banks pay extra interest on customer courting management. A robust and real-time credit card holder’s churn evaluation is vital and valuable for bankers to preserve credit cardholders. Much research has been observed that retaining an old customer is more than five times easier compared to gaining a new customer. Hence, this paper proposes a method to predict churns based on a bank dataset. In this work, “Synthetic Minority Oversampling Technique” (SMOTE) has been used for handling the imbalanced dataset. Credit card customer churn is predicted using random forest, k-nearest neighbor, and two boosting algorithms, XGBoost and CatBoost. Hyperparameter tuning using grid search has been used to increase the accuracy. The experimental result shows Catboost has achieved an accuracy of 97.85% and tends to do better than the other models.

Air pollution in cities has become an important topic due to its adverse effects on humans and air quality. The aim of this paper is to monitor particles less than 1 $$\upmu $$ m (PM $$_1$$ ), 2.5 $$\upmu $$ m (PM $$_{2.5}$$ ), 4 $$\upmu $$ m (PM $$_4$$ ), and 10 $$\upmu $$ m (PM $$_{10}$$ ). The concentration of particulate matter highly changed with location and time. Particulate matter in the air is considered as the primary pollutant, and it affects the environment band the risk of mortality and morbidity of respiratory disease. To address this issue, the research presents the design and development of the low-cost network for monitoring particulate matter using sensiron sensor (SPS30). These devices are equipped with LoRaWAN to test the low-power wide-area network coverage. The designed network contains the sensors connected to the ESP32 microcontroller towards the processing of LoRa modules (sensor nodes), which send data to the gateway using the frequency band, using the ‘The Things Network (TTN)’. The sensor collects the different particle matter in the air. The proposed network design has been implemented at St. Paul Street Auckland. The designed network system allows the users to access the online dashboard to test and monitor the concentration levels of particulate matter in the air.

The speech command recognition system based on deep image classification is the key that would tremendously promise to revolutionize research and development by overcoming the communication barrier between human and machine or computer. We are all aware of challenges in identifying the voice command in noise and variability in speed, pitch, and projection. This paper has developed an efficient and highly accurate speech command recognition for smart and effective speech processing applications like modern telecommunication. In particular, a novel convolutional neural network (CNN) is presented that works with a one-second audio clip consisting of one specific word including ten speech commands and other words labeled as “unknown,” and model implementations were operated in the noisy environment. The CNNs are structurally fully developed in such a way to recognize the speech commands with the utilization of deep learning (DL) for image classification concepts. Thus, this research used the concept of DL for image classification to translate the problem of speech command recognition into the image domain.

In this pandemic situation, there are many problems facing people from all over the world. For example, economic downgrades, family problems, illness, and health problems, etc. Indonesia is one of many countries facing this situation. With a record of being the fourth most populous country in the world, Indonesia easily creates a lot of problems in this pandemic situation. This pandemic situation has made the world aware of such problems in their lives. One of the important types of problems that arise from this awareness is the maintenance of physical health. Maintenance of physical health is very important for this pandemic situation. COVID-19 can really cause many diseases, and therefore, health maintenance is very important to get rid of this virus. Many media inform the public to take care of their physical health, but there are also many things that need to be considered and continue life as it is. Therefore, information explaining the importance of maintaining physical health is needed by people in Indonesia, and this study tries to review the importance of health maintenance. This data can be shown in of an app, where it is all currently done digitally, in conjunction also with COVID-19 pandemic that has resulted all people to become digitally literate. In addition, there is also the risk of access to meet other people if you have to consult a doctor, so making a physical health maintenance application during the COVID-19 pandemic will be very helpful. Various educations can be given to the public, such as educating the public that taking supplements/vitamins is important in preventing the COVID-19 virus. Consuming minerals and vitamins for endurance is indeed important during the current pandemic, because the body needs these substances to increase endurance. Apart from supplements, vitamin and mineral needs can also be obtained from various types of food. According to dietitian Maxine Smith, eating healthy foods is the best way to keep the immune system in balance, because the immune system relies heavily on nutritious food. This is because healthy foods contain various substances, including vitamins and minerals as immune support.

E-commerce has always been thought of as a fast-growing business, and even while online purchasing has not followed the same boom patterns as in the past, it is now being recognized for its potential. One of the current study topics in the field of textual content mining is sentiment evaluation. Digging normal language for conclusions and feelings is a dreary cycle. The best solution is to conduct a sentiment analysis. This gives crucial data to decision production in an assortment of fields. There are a variety of sentiment detection algorithms available, each of which has an impact on the quality of the final result. In this study, we investigate people’s feelings toward the services offered by e-commerce websites. Reviews, ratings, and emoticons are among the sentiments. The main purpose is to identify bogus reviews using user and product ids in order to purchase the best products and analyze which one is the best. For this, we deploy a mixed learning system that analyzes numerous service feedbacks. To track down the scores of each word, a text mining calculation is utilized. The sentiments are then divided into three categories: negative, positive, and neutral. It is possible to study a website to figure out whether there are any bogus reviews. Finally, anticipate the bogus reviews left by users. Only paid users can leave reviews, and duplicates are checked using the user id and booking id. Credible client criticism is incorporated when making item suggestions.

Encrypted communication technology has evolved as the network, and Internet applications have advanced. Malicious communication, on the other hand, employs encryption to bypass standard detection and security protection. The existing security prevention and detection technologies are unable to identify harmful communication that is encrypted. The growth of artificial intelligence (AI) in these days has enabled to employ machine learning (ML) as well as deep learning approaches to identify encrypted malicious communications without decryption, with remarkably precise detection outcomes. At this moment, research on detecting harmful encrypted traffic is mostly focused on analyzing the features of encrypted data and selecting neural network (NN) techniques. Hybrid ML is proposed in this study by merging two well-performing data mining algorithms with natural language processing tasks. Here, a new traffic flow detection method is performed by the hybrid ML technique. At first, the benchmark data is collected from public sources. The features are extracted using the convolutional layer of deep convolutional neural network (DCNN). Then, the weighted feature extraction is performed by grasshopper optimization algorithm (GOA). Employed the hybrid machine learning-based malicious detection with the “support vector machine (SVM) and neural network (NN)” is utilized in this model to detect the traffic affected by malicious activities, where the hidden neuron count of NN and kernel of SVM are tuning by the same GOA for increasing the accuracy and precision. This research provides findings from experiment, encouraging various researchers to develop the research as future work.

In line with the strong growth of the social data volume, a huge number of social contents are created in various forms every day. As a result, it is often difficult for users to find or keep up with current popular keywords. Many researchers are interested to analyze the social data and identify the current popular keywords and topics. Thus, these researches are more complex for profiling the target topics with their keywords. The main purpose of this paper is to identify the most trending keywords in social media based on profiling influencers. So, for achieving this objective. Our approach illustrates determining influential nodes on a social network graph. After that, due to the latent Dirichlet allocation model, we could profile the trending keywords based on influential posts content. In any case, our proposal is more relevant and optimal compared to the previous approaches.

Fire is an unpredictable dangerous event to the environment and the public. It may cause human and non-human belongings very rigorously. In recent days, many unpredictable events in the world are identified and controlled by computer vision-based machine learning approaches. Among these events, this looks forward to predicting the fire event and saving the human and non-human valuable belonging. Fire detection is a challenging task for the researchers to predict the root cause and alert the nearby ones. In this article, many algorithms have been addressed for fire detection methods from the past two decades that how the computer vision techniques have been grown day by day to predict the fire even. Even though, a formal survey on recent trends to fill the gaps is identified with the last decades. Countless fire detection strategies are developed through machine learning, deep learning, computer vision, etc., with images, video, and sensors. This article discussed a detailed survey that how the fire detection process becomes popular in the research area in recent days. In addition, a proposed new model is to rectify the issues in existing approaches.

An overwhelming number of disease surveys capture patient records of the severity of their illness indications that enable differentiation of patient ailment from typical indication. This paper aims to forecast illnesses in users based on their symptoms. To achieve our aim, we use the XGBoost Classifier, which aids in regulating the victim’s health condition then acquiring the symptoms to predict the condition. The data carries 250 variables and 140 corporal evaluation instances (diseases). In addition, we provide medicine for illness and gather the corresponding victim EH to summarize the diagnostic described by the NLTK.

Landmark recognition is a method that predicts landmark labels directly from image pixels. This model is designed to detect the proper landmark in a dataset of complex and challenging test images, allowing users to better understand and organize their photograph collections. A dataset containing historic monuments of India of around 100 classes is created. The dataset is divided into training, testing, and validation data. Data augmentation is performed on training and validation data in the dataset. ResNet50 algorithm is used here to train the dataset. A sample image is given to the ResNet50 model for prediction, and the model gives the top ten labels of the sample image. Consider two images for each label from the training dataset, and apply DELF on these images and sample image. The inliers are calculated for these images using the Ransac algorithm. Image label which has a good number of inliers with sample image is the landmark of that sample image.

The prices of agricultural commodities are constantly varying. It is affected by various factors such as weather conditions, plant diseases, labor charges, production, demand, and supply of agricultural products. The farmers face problems when the crop is not worth the price and when unaware of the marketing price. This makes it difficult for them to do agriculture. Future crop prices can be estimated using machine learning algorithms such as decision trees, logistic regression, and support vector machine (SVM). In this paper, we analyzed five supervised algorithms and chose SVM as it gave better results. The dataset contains the prices of agricultural commodities from various states and their nutrient value. At last, the XGBoost boosting algorithm is applied to improve the performance of the model.

The Industrial Revolution 4.0 has significantly altered how people think, live, and interact. This period has disrupted a wide range of human activities, not just in the sphere of technology, but also in other areas like economics, social relations, and politics. A mix of technologies that blend the physical, digital, and biological realms, together known as cyber-physical systems, characterizes the Industrial Revolution 4.0 age, thus requiring extensive action to help analyze and mitigate. Electronic payment is also a part of the Industrial Revolution 4.0, and it has so many impacts. In this research, we are going to see what it does to small businesses. This type of research will be using quantitative method with 43 samples from small businesses’ owners using questionnaire method. The results show that electronic payment has so much impact on small businesses such as more alternative payment options.

Over the past several years, the recommendation system has become one of the important aspects in our day to day life. Recommendation systems are used in various areas like YouTube, Amazon, etc. Collaborative filtering algorithms have been used in recommendation systems for a long time. They have been effective in resolving a number of issues with commercially available systems. Methods based on a user’s neighborhood have showed potential in forecasting user ratings. The aim of this paper is to design and evaluate ‘KNN algorithm and Collaborative Filtering algorithm’ for producing movie recommendations. The dataset used in this paper is ‘Movielens dataset’ which is downloaded from Kaggle. The system was implemented using ‘Python programming language’. Initially, compared different distance measures and then performed correlation. Based on the correlation value, developed system will recommend similar movies/users. Performances of both the developed algorithms were analyzed in terms of accuracy. Finally the result shows that the accuracy of the recommendations is very good and we will get more accurate movie recommendations based on the combination of “KNN algorithm and collaborative filtering algorithms”.

Internet of Things (IoT) is based on characteristics of the Wireless Sensor Network (WSN) in which nodes are dispersed for segregating data for different applications. In IoT, the sensor nodes typically possess heterogeneous properties. Among, few nodes have higher energy and even data aggregation functionality. Generally, in WSNs, efficient cluster-based routing is employed for data transmission. When Cluster Head (CH) fails, data sensed by the sensor nodes cannot be transmitted by the faulty CH. Consequently, the gateway or sink cannot recognize the data in the IoT sufficiently. Hence, processing information in this field was severely affected. This research contribution focuses on paired fault-tolerant cluster routing of the disjoint routes in a data flow graph and introduces a new approach called Free Poisson Law for solving this problem. The proposed Efficient Data Flow Graph Modeling using Free Poisson Law (EDFGM-FPL) algorithm has the aim of reducing latency, energy consumption as well as end-to-end delay, dissipated energy, functional complexity thereby achieving improved packet delivery ratio, throughput, and fault detection rate.

Automatic text summarization tasks play an important role in natural language processing. In this work, we introduce the single-document extractive summarization model based on clustering and word embedding. In the model, we use K-Means clustering to create the clusters on the large-scale dataset by using word embedding as the feature vector, then use these clusters to extract the most relevant sentences on the document to summarize. At first, we collected the articles on the Vietnamese online newspapers, cleaned them and built up the dataset with a total of 1,101,101 articles. After that, we applied our summarization model for the experimentation. The average time cost for summarizing one document in the test set is 6.22 ms, and the best F-Score of this model based on ROUGE-1, ROUGE-2, and ROUGE-L are 51.40, 16.15, and 29.18%.

The paper considers methods to improve the efficiency and reliability of the multi-agent system of knowledge representation and processing (MASKRP). An approach to the implementation of MASKRP control subsystems based on an exo-kernel operating system and specialized event-driven software modules is proposed. The tasks of effective distribution of software agents on the nodes of the computing system, optimization of Distributed Knowledge Base (DKB), formation of computing clusters with minimal information connectivity are solved. The architecture of the monitoring system is reviewed. The main results on the development of software for searching anomalies in the work of the MASKRP computing cluster equipment as well as on the determination of the optimal parameters for it to prevent false alarms are given.

For understanding the content of the image, a vital clue is the superimposed text in images. However, ineffective text extraction might be caused by the poor contrast and complicated background of the image. An effective text detection and recognition framework as of natural scene images utilizing Anopheles search with convolution neural network are proposed to mitigate such complexities. In this work, for detecting and recognizing the exact text from the image, different steps are undergone by the input images from the street view text (SVT) dataset. The precise text is extracted from the images. For comparing the results attained by the proposed AS-CNN with the preceding top-notch algorithms, SVT and publically accessible datasets are utilized in this research. The experimental outcomes denoted that the proposed AS-CNN exhibits promising results, effectively decreasing the overall computation cost and time.

For semantic video retrieval (from archives), alert production (live streams), and large levels applications like assessment mining and content reiterating, textual material in videos is an appealing index. The discovery and recognition of textual content are essential factors of similar systems, and this is the focus of our research. This case study talks about all types of the framework for text recognition and segmentation in video frames. We focus on cursive scripts, in particular, using English text as an example. Fine-tuning deep neural network-based item finders for the particular instance of text identification is used to recognize textual patches in video frames. Convolutional neural networks (CNNs) are used to identify the script of the discovered textual material, while an EngNet, a hybrid of CNNs, and long short-term memory (LSTM) (Rajesh Kanna and Santhi in Expert Syst Appl 194, 2022 [1]) systems is proposed for identification. Also being generated is a benchmark dataset with cursive writing and over 13,000 video frames. An F-proportion of 88.3% for identification and an F-proportion of 87% for acknowledgment was later a total arrangement of preliminaries.

An analysis of cyclostationary malware is introduced. The most important cyclostationary features used for network intrusion detection systems (NIDS) are then detected with a feature extractor algorithm, such as Boruta and principal component analysis (PCA). These feature patterns are classified to determine the most cyclostationary ones. In particular, this article shows the relevance of detecting cyclostationary malware for a NIDS by using legacy datasets, such as the KDD99 and NSL-KDD. This research has also used the UGRansome cyclostationary dataset intended to support research on anomaly detection. This dataset is subdivided into normal and abnormal classes of network threats. A comparative analysis based on random forest and support vector machine algorithms is undertaken, and the performance of Boruta and PCA was also evaluated. The research suggests the utilization of PCA in terms of extracting cyclostationary network feature patterns as a viable proposition compared to Boruta. The Internet Protocol (IP), malware financial damages, class C of IP addresses, and signature malware were also found to be the most cyclostationary feature pattern. The UGRansome dataset outperformed the KDD99 and NSL-KDD in terms of detecting signature malware with an accuracy of 99% using the random forest algorithm, while the support vector machine achieved 68%. This research proposes the UGRansome as a suitable choice to reduce the computational time of cyclostationary malware classification. Lastly, the research suggests the utilization of random forest to stratify and detect cyclostationary malware.

The aim of the article is to show the role of digital skills and literacy as a consequence of the processes of digitalization, but also as a prerequisite for conducting online learning in educational systems. The different dimensions of digitalization and the formation of digital skills by people of different ages and professions are analysed. The methodology is based on various surveys, including an online survey conducted by the authors of the article among the Bulgarian population on the state of digitalization and online learning. An important place in the study is occupied by online education in secondary and higher education in the situation of the COVID-19 pandemic. Views and assessments of the participants in the learning process and their readiness for mobile learning are shown. The main conclusion of the article is that in Bulgarian society there is a relatively good availability of digital skills, which are important for modern processes.

The perishable food production network is one of the most challenging areas in the food business; however, supply chain management is improving rapidly to satisfy the rigorous industry standards. With the global COVID-19 outbreak, food safety is having severe problems, especially in terms of customer purchase. On the other hand, people are more concerned with food quality, origin, and shipping regulations. In this context, a food discernibility framework is required for avoiding the over-changing of the stockpile area. The proposed model involves two main components such as information communication system and chemical-based access. In the grant delivery chain, the above listed tools are used to pinpoint, keep in sight, and keep under surveillance of the purchased food items. With the perishable food worldwide exchanges, client conduct, framework proficiency, unwavering quality, and different elements that share the data in the production network have become problematic as the e-commerce industry has grown. Block chain along with other distributed ledger technologies (DLTs) are fore casted in many organizations due to the allocation of storing the dataset that can be switched between organizations that don’t trust each other. ML’s learning capabilities may be integrated with blockchain-based apps to boost their capabilities and make them smarter. The disseminated record and time upgrade thrives a reasonable information sharing by using ML security methods. Food-trackable order collects data on food details and identifies food through supply chain activities by using the proposed blockchain KNN method in machine learning domain. The proposed research study will review the existing encryption technologies as well as other machine learning methods to show how the proposed blockchain with machine learning algorithm enhances efficiency.

The onset of blockchain generation within Bitcoin has generated significant interest by showing an opportunity to eliminate the middle ground need and transform communication between people and machines by increasing trust. Initially restricted to the integrated currency domain, people began to see the power of a generation beyond just the cryptocurrencies, which brought the acceptance of the blockchain era to erase the world’s problems. One such situation is problems for e-governance companies in other areas of the public sphere. In the scope of this thesis, we have specifically addressed the issues within the traditional property registration system. This thesis discusses the new design and architecture for real estate transactions and implements it using a blockchain-based solution and addressing issues including record integrity, privacy, and most importantly the lack of common platforms among concerned government organizations. The advent of the blockchain era led to the creation of blockchain-enabled platforms like Hyperledger Fabric. It is one of the most popular open-source permissioned blockchain frameworks, created and supported by the Linux Foundation and IBM, used in many industrial scenarios. So, it is used to create a network with one ordering organization with an ordering node and one peer organization with two peer nodes to prove the concept. Chaincode similar to Ethereum’s smart contract contains the logic to perform all operations and modify the ledger data. All the methods of chaincode are accessed using the Fabric gateway in the web application to perform various operations.

Voice is one among the important services provided by any mobile cellular network. The deployment of 5G that is the advanced mobile cellular communication technology will be more important than ever for the world, through various types of use cases. A 5G smartphone will not connect to a mobile network unless voice support is available, so enabling this is a must. Voice service requires end-to-end 5G network support to enable the high-quality voice service experience for mobile devices. So, we need to consider the whole network chain with IP Multimedia Subsystem (IMS), 5G core, and 5G radio access network (RAN)s. To provide seamless voice handover when users move between cells and access technologies, the network must also handle interworking toward 4G, 3G or 2G in case the users move out of 5G coverage. Evolved packet system (EPS) will be helpful for service providers when they are migrating from 4G to 5G. Circuit Switched FallBack (CSFB) will be helpful for service providers when they are migrating from 2G/3G to 4G. Once the call is set for voice communication if the mobile host (MH) moves from one cell in the wireless cellular network to another and finds no free channel in that cell, then the call will be dropped. To address this issue in this paper, we have presented an optimal channel reservation (OCR) policy. It reserves channels in a cell according to the target handoff call dropping probability (HCDP). So, it minimizes new call blocking probability (NCBP) and keeps HCDP below the target. We have applied this policy to Global System for Mobile Communication (GSM) 1800 system and observed that HCDP is below the target and NCBP is minimum.

In the current state of the pandemic, all sectors of human life must change because of the health protocols that must be implemented. All sectors ranging from the economic sector, law, and education experienced obstacles in its implementation. Of course, because this activity will be the spread of the COVID-19 outbreak cluster. In the field of education itself, in 2020 the government stipulates that the implementation of education in Indonesia is carried out using the distance learning model. However, these changes must of course be supported by technological advances in order to be achieved. Technologies such as zoom and other communication tools support this research. In addition, technology that is no less important is big data analysis in decision making in educational institutions, which is also needed to improve distance learning. But is distance learning effective today? And how does the use of big data analysis affect the effectiveness of the distance learning process? To prove it, we chose one of the universities that have used big data analysis in Private University in Jakarta as the sample to be analyzed using quantitative research methods with a Likert-shaped survey distribution. The results of this survey are that according to respondent’s distance learning is currently effective, and indeed there is an effect of increasing the effectiveness of distance learning by using big data analysis to assist decision making.

Climate change and natural disasters are especially dangerous to coastal landscapes. Natural and man-made hazards like erosion sedimentation, rising sea levels, and tidal flooding are wreaking havoc on coastal areas. The mapping and detection of coastlines are critical for safe navigation, environmental conservation, and long-term coastal development. The model is composed of a method for extracting coastline from sentinel-2-l2a satellite images obtained from the sentinelhub website. The model uses the Gaussian blur module to reduce noise. Using k-means clustering, the image is then segmented as land and water. An edge detection algorithm identifies the boundary between land and water. One of many edge detection algorithms, Canny edge detector, applies a multi-level algorithm to detect edges in images. Percentage change is calculated by comparing window to window of edge detected images.

The increment of the world population is the main reason behind the emission of greenhouse gasses like methane and carbon dioxide which affect temperature and humidity. To identify this problem, we implemented an IoT system using an ESP32 microcontroller and the necessary sensor to measure data from the outdoor environment and deploy our system to monitor the outdoor temperature and humidity quality due to increasing the level of methane and carbon dioxide. In the next step, the measured data was sent into the ThingSpeak cloud server and later was made available for visualization. This research work seeks to estimate the level of methane and carbon dioxide emissions which has a great impact on increasing temperature and humidity and also highlights some complications to integrating modern sensors based on the emergence of the Internet of things (IoT) concept. However, the data preserved in the cloud server can be used for further analysis to find other impacts of gasses in our environment.

A new intelligent monitoring model is developed by us for determining gender and nationality from frontal picture candidates utilizing the face area based on deep learning. Face recognition is influenced by a number of characteristics, including picture quality, illumination, rotation angle, blockage, and facial expression. As a result, we must first recognize an input image before converting it to a genuine input. Nationality is the most well distinguishing characteristic that is applied in every nation, and it is also important to secure authentication. Image detection is crucial in this case. Then, we can determine the facial shape, gender, and nationality of the candidate image. In the end, we return the result based on the distance comparison using the use of a library to measure the sample. There were significant discrepancies across photographs while measuring samples based on their gender and facial features. The photos used in the input must be the same as those used in the output. Picture of a frontal face with clean lighting and no blemishes at every angle of rotation. The model may be used by ordinary people, models, celebrities, actors, and others. In the end, computer can tell nationality by looking at a picture of a person’s face.

Delay Tolerant Networks, the heterogeneous wireless network, are having the ability to enable communication in intermittent connectivity. But, the network suffers from poor performance due to low contact opportunity between nodes. Contact opportunity between nodes must be increased to improve the performance of the network. For that we can deploy static relay nodes also called as throwboxes in optimistic places so that more number of nodes can communicate by that node. Hence, performance of the network can be improved. In this paper, we proposed a deployment technique using grey wolf optimization (GWO), a metaheuristic technique to increase delivery ratio in minimum delay. In this paper, the objective is to find the optimal places using GWO hence, delivery ratio of the network can be increased with least end-to-end delay. For simulation, we have used ONE simulator to find the result and comparison with previously proposed deployment schemes in DTN. By analysing results, we can say that proposed deployment technique performs better as compared to existing deployment technique.

Forests are considered synonyms for abundance on our planet. They uphold the lifecycle of a diversity of creatures, including mankind. Destruction of such forests due to environmental hazards like forest fires is disastrous and leads to loss of economy, wildlife, property, and people. It endangers everything in its vicinity. Sadly, the presence of flora and fauna only increase the fire spread capability and speed. Early detection of these forest fires can help control the spread and protect the nearby areas from the damage caused. This research paper aims at predicting the occurrence of forest fires using machine learning and deep learning techniques. The idea is to apply multiple algorithms to the data and perform comparative analysis to find the best-performing model. The best performance is obtained by the decision tree model for this work. It gave an accuracy of 79.6% and a recall score of 0.90. This model was then implemented on front-end WebUI using the flask and pickle modules in Python. The front-end Website returns the probability that a forest fire occurs for a set of inputs given by the user. This implementation is done using the PyCharm IDE.

To ensure peace and safety of the country, military forces are inevitable for the country’s protection. India has the second largest military force in the worlds with the strength of 1.4 million active troops. The army people lose their valuable lives due to lack of technical infrastructure to provide status about the injured soldiers. Many lose their lives and become disabilities because of injuries rather than battlefield. Nowadays, there are vast amount of electronic gadgets to fight against the enemies and secure the nation in the border. But in some adverse conditions like bad weather or due to remote location of soldiers, it becomes impossible to interact with them and monitor their health conditions of militants. In this work, we are developing a modern wearable smart thong entrenched with system of system facility and mobile ad hoc network. It has the special feature of continuously monitoring the health condition of troops. It also continuously records the health parameters and environmental parameters of the military warriors so that valuable lives of soldiers which is significant asset to our country can be rescued at right time.

IoT devices have become an integral part of our lives. The world has witnessed an exponential growth in the number of IoT devices. Managing these devices and the data generated by them has become very crucial. Data security and users’ privacy are becoming more difficult as the number of devices grows. Blockchain, the technology behind Bitcoin, is known for data security and managing and efficiently maintaining huge amounts of data. Blockchain stores data in a chronological manner and in an immutable way. Integration of blockchain in IoT infrastructure has many advantages. This paper discusses various applications and challenges in blockchain. It highlights the adoption of blockchain in IoT infrastructure and reviews recent papers in this field.

As image sensor electronics have progressed, and hyperspectral images have been used in a wide range of applications. In terms of recognizing the classes, a lot of research work has been done to extract useful information from the available unstructured knowledge database. The use of spectral and geographical datatypes in images can improve the classification precision. To improve the accuracy of the energetic-spectral snap analysis, combining dimensional and spectral data is a good idea. This research study examines the history of dimensional facts based on energetic-spectral image classification designs by using prepared and semi-directed classifiers to classify detached sensing images with particularized class labels. Long-term data are removed if the features are protensive. To increase the veracity of classification, the extracted traits are prepared by utilizing multiple classifiers. The preparation and sinking balance towards the loss function have been reused to train the classifiers. To avoid local minima, the preparation is approved by utilizing various tiers for accompanying the extra balancing limits. To overcome the risk of establishing a contradictory validity image, each detached perceiving image is classified throughout the experiment stage. Exploratory discoveries demand higher validity in-class criteria than other advanced directed classifier techniques.

A finite state machine, commonly known as a finite state automaton, is a model of computation based on a theoretical machine and is composed of one or more states. It enables the switch between multiple possible states and modification of the behaviour according to the state. One of the widely used applications of finite automata is in the video gaming industry. Game designing is the methodology of developing the content and rules of a game in the initial stages. It involves designing the characters, rules, environment, action, and rewards. For example, Super Mario is a game series based on the fictional plumber Mario created by Nintendo. The fundamental purpose of the game is to move the character, Mario and complete as many stages as possible. The secondary goal is killing the enemies and collecting coins or additional items. This project involves Mario transforming his states and behaviour based on the events that have occurred using finite automata. This paper deals with implementing the Super Mario Game Design using Python Programming Language and JFLAP, a widespread open-source software tool used in the Formal Language and Automata courses.

Spam email is a massive issue that bothers and consumes receivers’ time and effort. Because of its effectiveness in identifying mail as wanted or unwanted, machine learning approaches have become a popular technique in spam detection. Current spam detection methods, on the other side, typically have low detection performance and are incapable of handling high-dimensional information easily. As a result, a unique spam detection approach that combines an improved moth flame optimization algorithm and a logistic regression classification model was proposed in this paper. The research evidence on two accessible datasets (CSDMC2010, Enron) indicates that the suggested methodology can tackle high-dimensional data due to its very powerful local and global search skills. The suggested technique was evaluated for spam detection accuracy to that of logistic regression, naive Bayes classifiers, and support vector machine, as well as the performance of earlier research’ that includes state-of-the-art approaches. In terms of classification performance, the suggested methodology outperforms the other spam detection algorithms examined in this work.

World Wide Web has become a universal environment for human interface, collaboration, communication, data storage and data sharing. Information retrieval is a technique to understand the text in the web pages. The purpose of Semantic Web is to allow the machines to advance knowledge itself by recognizing its meaning. The essential intention of this research is to predict variation in the directional behaviour of the digital library user, based on their recommended data. The significance of this research lies in enlightening the user behaviour in digital library using semantic-driven approach. An important contribution of this research work corresponds to retrieve accurate data from websites by using semantically enhanced various algorithms. This research concerns on identifying educational digital library web user’s behaviour. The user behaviour unstructured data are pre-processed and semantically extracted using suggested weighted TF-IDF normalization method. Then the extracted features are selected using improved Ensemble-based FS method. It improves the performance and reliability of classification by eliminating unnecessary and superfluous features from the extracted user behaviour datasets.

The development technology at this time had a complex impact on human life, and education was no exception. The education world at this time has used learning technology media to improve its quality, including high school education. In order for this utilization process to operate effectively, various operations of the learning system are of course required. As we all know, the learning system used to only focus on the teacher, who would explain the material directly to the students. However, with the development of learning technology media, students are expected to become more active in the learning process. The purpose of this study was to understand how learning technology media can help the learning process in high school. The results suggest that the development of learning media technology can stimulate students’ creativity, as well as assist the learning process, because apart from being a learning medium, this technology can also be used as a tool and media for storing and distributing material. So that the development of learning technology media can improve the quality of education in high school.

Microelectromechanical systems (MEMS) technology is extensively used for making high-performance sensors and actuators in Internet of Things. The ubiquitous advantages and potential applications of MEMS sensors compared to conventional ones paved the way for implementing the diverse micro-sensing technologies in Internet of Things (IoT). Vibrating plate-based MEMS resonant sensors are the most commonly used small structural elements due to its low mass and high-quality factors. The downsizing of devices for achieving minute sensors used in IoT leads to application of higher-order theories. The requirement of enhanced quality factors necessitates the study of characteristics affecting the quality factor and associated energy dissipation. The current analysis explores the significance of mode switching on energy loss which controls the most decisive performance measure for sensors used in IoT applications. The investigation of impact of mode switching on thermoelastic energy dissipation with different structural materials for a rectangular plate-based resonator is included in the study. The energy dissipation is verified to be slightly diminished when operated in higher modes, and the quantification of the impact is accurately done by calculating the percentage reduction in energy losses. While considering the thermoelastic energy dissipation of microplate-based resonators, the influence of mode switching on diamond-based plate resonators is verified to be the most prominent according to our findings. The result analysis of the proposed work helps the engineers for designing resonators with diamond as the structural material operating in higher vibrating modes for IoT and 5G applications.

Depression is an extremely serious illness of humans which causes constant mood swings and feelings of sadness. Nowadays, it is considered to be a deadly disorder in the world. At present, everyone from young to old is suffering from depression but most of them do not have the right idea about their mental state. Everyone needs to have a proper idea about their mental state. We will detect depression through a machine learning-based detection approach. Talking with psychologists and depressed people, we find some factors that are related to becoming depressed, and depending on those factors, information is collected from both depressed and non-depressed people. After applying preprocessing techniques, a processed dataset was created finally. Then, feature selection techniques were used. We applied eight machine learning algorithms and two feature selection methods to our dataset. We used k-nearest neighbor (k-NN), decision tree (DT), linear discriminant analysis (LDA), adaptive boosting (AB), support vector machine (SVM), naive Bayes (NB), random forest (RF), and logistic regression (LR) classifier. In our work, the RF classifier gave the best performance based on accuracy and the accuracy of the RF classifier was 96.00%.

Vietnam is among the top countries in the world at risk of natural hazards so that efficient real-time environment monitoring is becoming essential. The continuous development in information and communication technology is inspiring the development of the smart monitoring systems for environmental management and protection. This paper presents an energy efficient, long range sensor system for Internet of Things (IoT)-based smart environment monitoring and early warning. The proposed system combines the novel energy efficient temperature. Beat sensors integrated with the LoRaWAN communication protocol and web interface. The experimental results are achieved to clarify the efficiency of the proposed sensor system and its potential applications in the real systems.

Categorical search and category-wise book recommendation are two common tasks for online booksellers. But for a machine to understand this category from a given text is still challenging work, where machine learning is a widely used tool at present. Though in the English language, with the availability of rich datasets and corpus, machine learning-based categorization and recommendation have reached a standard level, in the Bengali language, to reach the standard, still needs a long way to go. One key reason is the lack of availability of a rich Bengali dataset. The aim of this research was to make a dataset first for the book’s genre identification from its given summary and to explore which supervised classifier performed best on that dataset for classifying the genres. Before that, we performed several essential preprocessing steps essential to prepare our dataset fit for the algorithms. Six machine learning classifiers were applied to the dataset, and it was observed that Naive Bayes performed best with an accuracy of 68% followed by XGB with an accuracy of 67%.

The pandemic known as COVID-19 has engulfed the world, and brought with it a huge set of impacts on economic growth, especially in the MSME sector for a country, especially Indonesia. Because this pandemic has hit various MSME sectors in Indonesia, especially the tourism sector, it is necessary to get more attention in order to get fast handling. Therefore, this study will discuss the impacts of COVID-19 MSMEs that operate in the tourism sector, what some tourist-heavy regions in Indonesia have been doing thus far to aid in overcoming them, and what the government can do to overcome these impacts. The purpose of this study is to help find solutions to these problems. Our research employs a qualitative research method with a descriptive type of research. The results of the research show that there are several impacts that the pandemic had on MSMEs in the tourism sector, including the lack of income generated because the source of income comes from tourist visits and a large drop in earnings and a large increase in MSME bankruptcy across the nation. Most of this occurred because the presence of COVID-19 prevents tourists from visiting. Each region has also enacted their own solutions, such as health protocols in Lampung and Kupang, and tourist-attracting marketing for Bali. The research also shows that the ministry of tourism has enacted several policies to countermeasure this issue, such as a change in how tourist-based businesses are run, safety protocols and certifications, and regulations. Our solution consists of contactless technological solutions as alternatives to travel and enactment of previously discussed government policies.

Log generation is a continuous process that generates large amounts of log data in various forms and rates that may be analysed to acquire valuable insights.Various open-source and commercial alternatives for integrating log acquisition, storage, retrieval, and analytic have been created in the light of the current Big Data era and the importance of log analysis. Taking into consideration benefits mentioned in literature about Spark and ELK for log analysis, unifying Spark and Elastic Stack capabilities with other open-source services and platforms helpful for log analysis can really prove beneficial for developing an efficient framework for scalable log management. In this study, we present auto-scalable Big Data enabled log analytic framework integrating Apache Flume, Apache Kafka, ELK Stack, and Spark. By combining the functionality of all of the platforms mentioned, the framework seeks to deliver a platform that integrates online and offline Big Data processing methods while also enabling real-time analytic. We also deployed crucial components of the modelled framework on a horizontally scalable Kubernetes cluster implemented over AWS cloud, anticipating performance advantages from the dynamic scalability given through implementation design. Furthermore, the performance of the modelled novel framework is analysed using various search and analytic criteria both before and after deployment over AWS cloud. It was discovered that the novel framework outperformed in terms of several search and analytic criteria used to evaluate performance. We further critically examined the implications of this cloud-based deployment on the framework’s performance.

The cognitive impairments among elderly peoples make their lifestyle more complex, and it is extremely essential to accurately discover patients with mild cognitive impairment stage as it may or may not progress to Alzheimer’s disease. Many machine learning models have developed to discover the prediction of Alzheimer’s disease, but they are constrained with imbalanced medical dataset with the disease case records. Thus, this paper concentrates on developing the robust model which handles the imbalanced data when there are a smaller number of instances with case label of dementia. In this work, neuro fuzzy symbolic reasoning is developed by adapting case-based reasoning to retrieve similar pattern of matched records with the new instance as input query to determine whether it suffers from dementia or not. The case-based reasoning is empowered by applying cuckoo search model, which involves searching for matching records by applying levy flight strategy to define unknown instances and their corresponding matching instance labels, and this information is given as input to the fuzzy artificial neural network along with training dataset to enhance its prediction accuracy. The FANN with the additional knowledge obtained from cuckoo search-enhanced case-based reasoning during training phase handles uncertainty and produces more accurate results for Alzheimer’s prediction compared to the other traditional models of classification. During testing phase, the new cases are predicted and if the output is correct, then it is added and maintained in the case history for future references. The simulation results proved the prominence of cuckoo search-enabled symbolic reasoning-based Fuzzy Neuro Classifier by applying this strategy on ADNI dataset.

Nowadays, the number of vehicles on the road is steadily increasing, and this raises the risk of road accidents. One of the reasons of this regrettable situation is violation of traffic rules. Overloaded cars are one example of a traffic rule violation. This research study proposes a hybrid approach for detecting different types of violations and violators. The proposed model will detect the overloaded vehicles and the number plate of the offenders by using the deep neural network model called YOLOv4. Additionally, Tesseract is used to recognize characters in the LSTM number plate. Finally, the proposed model has delivered a satisfactory result.

Nowadays, an important topic that is considered a lot is how to integrate Machine Learning (ML) to cloud resources management. In this study, our goal is to explore the most important cloud resources management issues that have been combined with ML and which present many promising results. To accomplish this, we used chronological charts based on keywords that we considered important and tried to answer the question: is ML suitable for resources management problems in the cloud? Furthermore, a short discussion takes place on the data that are available and the open challenges on it. A big collection of researches is used to make sensible comparisons between the ML techniques that are used in the different kind of cloud resources management fields and we propose the most suitable ML model for each field.

Correction to: Chapter “A Delicate Authentication Mechanism for IoT Devices with Lower Overhead Issues” in: S. Smys et al. (eds.), Computer Networks and Inventive Communication Technologies, Lecture Notes on Data Engineering and Communications Technologies 141, https://doi.org/10.1007/978-981-19-3035-5_7