Computational Science and Technology

Most data in the world is in form of text. Therefore, we can say text stores large amount of the knowledge of human beings. Extracting useful knowledge from text, however, is not a simple task. In this paper, we present a complete pipeline to extract knowledge from paragraph. This pipeline combines state-of-the-art systems in order to yield optimal results. There are some other Knowledge Graphs such as Google Knowledge Graph, YAGO, or DBpedia. Most of the data in these Knowledge Graphs is in English. On the other hand, the results from our system is used to build a new Knowledge Graph in Vietnamese of Vietnam Tourism. We use the rich resources language like English to process a low resources language like Vietnamese. We utilize the NLP tools of English such as Google translate, Stanford parser, Co-referencing, ClausIE, MinIE. We develop Google Search to find the text describing the entities in the Internet. This text is in Vietnamese. Then, we translate the Vietnamese text into English text and use English NLP tools to extract triples. Finally, we translate the triples back into Vietnamese and build the knowledge graph of Vietnam tourism. We conduct experiment and discover the advantages and disadvantages of our method.

Technology Adoption Model is used in various technology fields to understand and predict users’ intentions and behaviors. However, the Technology Adoption Model used in Social Media, which explains users’ intentions and behaviors needs to be investigated. Nevertheless, there is little understanding of users’ intentions and behaviors towards visual information, which plays an important role in effective communication. This study reviewed a considerable amount of past studies on the use of the technology adoption model by users’ online social media behaviors towards visual information. Based on the literature survey from FOUR (4) databases; ACM, IEEE, Scopus, and Science Direct; TWELVE (12) articles have been reviewed. The study found that Uses and Gratifications Theory (UGT) is the most adopted model due to the motivation mechanism applied. Most importantly, the review managed to discuss the models, factors, visual information, and methods in relation to users’ social media intention and behavior. An Integrated Adoption Model could be developed to examine the consequences of the technology adopted to create a holistic understanding of how technology influences the users’ intentions and behaviors towards visual information. This is one of the recommendations presented at the end of this research for the reference of future scholars.

Although there is a variety of technology available in 21st century, the way classes and lesson being conducted are still mostly remaining the same which is teaching via one-way communication. One-way communication teaching process surfastces issues like lack of interaction where minimal discussion is going on during teaching session, Since the interaction between teachers and students are fairly poor, limited classroom activity can be expected such as “Question and Answer” which lead to the increment of boredom in classes and lessons among students. As the teaching process is leaning towards instructional, students’ feedback is usually being overlooked. Hence, since the availability of educational technology is getting more common nowadays, the integration of technology in classroom is encouraged changing the teaching and learning environment including knowledge delivery method from teacher perspective. Introducing mobile interactive system allows teachers deliver their knowledge differently, however, they are required to adapt and get familiar with the educational technology for improving teaching experience. Therefore, Technological Pedagogical Content Knowledge (TPACK) framework is applied in this paper to assess teachers using the seven elements of the framework. This paper is aimed to address the issues faced by conventional classroom and identify the effectiveness of teachers conducting classes using educational technology with the application of TPACK framework. Proposed TPACK framework is formed with the integration of three elements which are teachers’ efficiency, students’ performance and students’ engagement representing the outcome of current TPACK framework.

The ripeness of the farm-able palm fruits is an important factor in the production of quality palm oil. The work presented is an image processing implementation in the palm oil industry to eliminate human errors in the judgment of the ripeness of palm fruit bunches as well as to introduce automation. Various techniques were employed to obtain data from the images provided for the data mining process. The features used are the colour of the palm fruit bunches and the amount of edges representing visible leaves in the palm fruit bunches, indicating empty sockets. The project is able to achieve an accuracy of up to 79.11%.

In this research, we look at the notions of objectivity and subjectivity and create word embeddings from them for the purpose of sentiment analysis. We created word vectors from two datasets, the Wikipedia English Dataset for objectivity and the Amazon Product Reviews Data dataset for subjectivity. A model incorporating an Attention Mechanism was proposed. The proposed Attention model was compared to Logistic Regression, Linear Support Vector Classification models, and the former was able to achieve the highest accuracy with large enough data through augmentation. In the case of objectivity and subjectivity, models trained with the objectivity word embeddings performed worse than their counterpart. However, when compared to the BERT model, a model also with Attention Mechanism but has its own word embedding technique, the BERT model achieved higher accuracy even though model training was performed with only transfer learning.

This paper proposes a conceptual architectural design of Question-Answering (QA) system that can solve “counting” problem. Counting problem is the inability of QA system to produce numerical answer based on retrieved rationale (in text passage) containing list of items. For example, consider “How many items are on sale?” as question and “Currently shampoo, soap and conditioner are on sale” as retrieved rationale from text passage. Normally, system will produce “shampoo, soap and conditioner” as an answer while the ground truth answer is “three”. In other words, system is simply unable to perform the counting process needed in order to correctly answer such questions. To solve this problem, QA system architecture with following components is proposed: (1) A classifier to determine if given question requires a counting answer, (2) A classifier to determine if current system’s answer is not numeric, and (3) A counting method to produce numerical answer based on given rationale. Despite looking like a whole system, the proposed architecture is actually a modular system whereby each component can operate independently (allowing each component to be separately implemented by other systems). In essence, this paper intents to demonstrate a general idea of how the defined problem can be solved using a modular system, that hopefully also opens up more flexible enhancements in the future.

Patient registration is an essential process in every clinic and hospital before services are provided to patients. Usually, patient’s identity card or fingerprint (through a fingerprint scanner) will be requested for identity authentication in order to retrieve medical records of the patient. However, the current global health crisis of COVID-19 pandemic is raising concern on the hygiene and safety of sharing objects or touching surfaces. Same worry is also occurred towards the patient registration interaction process; further, hospitals and clinics are classified as high risk premises. Therefore, a contactless patient authentication for registration using face recognition technology is proposed in this work. In this system, a face is scanned and processed. If the face exists in the database indicating that the subject is an established patient, the patient’s records will be retrieved. Else, a new patient registration will be performed to register a new account. The efficiency of the system is assessed using our self-collected database. Empirical results show that the proposed system is able to attain 94% accuracy. But, an inferior performance is obtained, especially dealing with makeup variation.

Pattern recognition is a mature but exciting and fast-developing field concerning computer vision, image processing, shape drawing, and text analysis. The shape pattern can be recognised easily using this technique. Some children have difficulties in identifying shapes. Therefore, through shape drawing exercises, it helps children to understand better. This project aims to develop a mobile application, assisting children in practising drawing using pattern recognition. It can identify shape types by matching the shape pattern with the given input. It will classify the information provided based on the feature extraction using the Freeman Chain Code. Then, each shape pattern is recognised using regular expression tools. Functionality testing has been conducted on this application with an accuracy of 80%. The application will encourage children to draw more by giving feedback on the exercise that they do. It may assist children in learning a new, better way of drawing shape accurately while improving children’s fine motor skills.

The advancement of technology nowadays provides opportunities in education including the adoption of blended learning. The objective of this project was to examine Information Technology students’ preferences on blended learning. Findings found that most students preferred to use it in Science subjects/courses, both inside the class and outside the classroom, and learn via online with a blended learning method as revision after the class. However, students of low and medium math anxious students had no significant difference in their preferences. It is hoped that the findings of the project will help researchers to customise the incorporation of blended learning tools in students learning as well help educators in planning and adopting blended learning, and benefit students in their learning. The system developers also should consider designing and develop more suitable systems/apps for students in learning Language subjects.

In this paper, we investigate two relatively new optimization algorithms in facial recognition, the grasshopper optimization algorithm (GOA) and binary dragonfly algorithm (BDA) which had the best performance out of 13 optimization algorithms that were compared. We investigate the effectiveness of both optimization algorithms alongside two classifiers, k-nearest neighbor (KNN) and support vector machine (SVM). Performance evaluation of the four combinations, BDA-KNN, BDA-SVM, GOA-KNN and GOA-SVM, indicate near-ideal recognition rates, with the GOA variants slightly outperforming their BDA counterparts. When compared to other recently proposed facial recognition approaches, the proposed algorithms depict improved accuracy.

The strategic planning is important in bank financial management. The banks and financial institutions have to achieve multiple goals in optimizing asset, liability, equity, earnings, profit and optimum management item. The subjective judgment in assigning weight of the goals is a drawback in financial management of the banks since it may cause inconsistent judgment. In addition, there are no comprehensive studies done on comparison among the banks for benchmarking based on the past studies in optimizing the financial management. Hence, this paper aims to improve the bank financial management by proposing a two-stage goal programming (GP) model to compare and optimize the bank financial management. The proposed model is developed based on entropy method in determining the weight of the goal at the first stage before optimizing the financial management with GP model at the second stage. Four listed banks in Malaysia are investigated in this study. The results indicate that the goal for asset, equity and optimum management item have been achieved by all banks. Furthermore, the target value of asset, equity, earning and profit can be increased according to the optimal solution of the proposed model. The significance of this paper is to provide insights to the banks for further improvement based on the optimal solution of the proposed model.

The study of human mortality is growing in Malaysia, as accurate mortality rates are classified important especially for social policy planning. This research aims at evaluating the performance of three selected mortality forecasting models, namely the Lee-Carter, CBD and M8 model in which the two latter models are from Cairns, Blake and Dowd. We applied the Malaysian central death rates and the number of mid-year exposures to the models and estimate the goodness of fits of all models using the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC). In addition, the 20-year out-samples forecast errors are estimated from 1999 to 2019 using the Root Mean Square Errors (RMSEs) and the Mean Absolute Percentage Errors (MAPEs). The findings of this study suggest that the M8 model is the best model for fitting Malaysian mortality data with minimum AIC and BIC values, and by far the most accurate model with the lowest out-sample errors, particularly for higher age category.

Learning styles, cognitive traits, personality, and learning preferences can vary greatly. That is why there is a great variety in how people receive and process information. Personalizing learning materials according to learner’s learning styles could enhance learner’s learning motivation and lead to better learning performance. This paper examines the relationship between learner’s learning styles and learning performance by proposing three different sets of documentation to test the relationship between the two learning styles of Felder-Silverman and learning performance. To test the proposed documentations and hypotheses, 182 participants in Multimedia University, Cyberjaya, Malaysia answered the Index of Learning Styles (ILS) questionnaire by Felder-Silverman and participated in a documentation experiment in Python programming. The data gathered was analysed using statistical Chi-square test. The results showed that learning performance was enhanced when the documentation was provided in a learning style that matched the subject’s learning style. The confirmed personalised learning styles model can be beneficial to teachers and e-learning recommendation systems when they provide students with materials that are personalised.

With recent pandemic, many students cannot join the class in physical classroom. The needs for e-learning and self-assessment become more salient than before. The teaching mode has been changing from teacher-centered to student-centered method. E-learning environment is practically a highly essential software application in the education field. However, programming-specific functionalities are hardly to be found on most of the general-purpose learning platforms, which may be unwieldy and unnecessarily complex to instructors and students in the programming learning process. This research aims to design a self-assessment model for a better support of programming e-learning, especially with exist of mandatory programming-specific functionalities. It’s believed that student background and effort have close correlation with their programming performance. More data to verify the correlations associated with positive learning outcome. In this research, we highlight the relationship between student background and student performance levels for introducing personalised self-assessment sets for students to learn programming. We propose and discuss Language, Education, Achievement, and Programming (LEAP) and Programming Learning Assessment (PLA) models to fill in the gap between the background knowledge and student competencies. To measure the correlation between proposed models and student performance, an experiment that involves 65 respondents was conducted. The data was analysed with structured and statistical approaches. Preliminary study shows that there are multivariate effects of the English fluency on PLA model. With the increasing demands of IT and software development skills, this research will help in motivating and encouraging more people to learn programming.

This paper introduces the design and functionality of a university academic advisor chatbot, which leverages on the result of a prediction model to predict students’ academic performance, to do early intervention to assist students who may need academic guidance. The prediction model is based on students’ attendance and scores of formative assessments to predict the score of the final summative assessment using a suitable machine learning algorithm. Scikit-learn library using Python will be used in this research to run the machine learning algorithms. The chatbot will be developed using Dialogflow which is integrated with one of the text messaging apps and established connection to a database. The database stores students’ attendance, scores of formative assessments, scores of final summative assessments and the status of students whom the chatbot has reached out to. This research aims to reduce the workload of lecturers to reach out to every student who is predicted to have problems in their academic studies and at the same time, be able to assist students using a chatbot.

Over the past few years, next-generation sequencing has become an invaluable technology for numerous applications in the field of genomics. The success of these applications are dependent on the performance of each phase in the genomic sequence pipeline, which starts with read mapping. However, read mapping is computationally intensive since it requires mapping billions of reads to numerous locations in a large reference genome. Building a q-gram index hash table has proven to be an efficient alternative to reduce the repetitive scanning of the reference during the verification step. A q-gram index hash table stores the locations of each q-gram in the reference genome. To accelerate the process of building this data structure and to exploit the multi-core architecture, instructions can be executed in parallel and distributed to multiple CPU cores. This paper performs a comparison analysis between the sequential and multiprocessing implementation of the index build time of the three methods for building a q-gram index hash table. The implementation results show that all multiprocessing versions are faster than sequential ones, with speedups ranging from 1.53 to 2.57. Although the open addressing method yields the fastest index build time, the best speedup is achieved by the minimizer-based method.

Rhythm games are music-themed games that challenge players’ sense of rhythm and reaction skills. One such popular rhythm-based video game is Dance Dance Revolution, where players perform steps on a dance platform that is synchronized with music as directed by on-screen step charts. An issue that exists, not just in Dance Dance Revolution, but in rhythm games in general is the estimation of a chart’s difficulty level. While many methods and studies exist in generating and predicting chart attributes, there is no clear methodology existing in determining the optimal difficulty of a given chart. This paper aims to address the aforementioned issue in the game of Dance Dance Revolution by proposing a methodology that involves extracting patterns and common attributes in step charts that enable more accuracy in determining a chart’s difficulty level. The resulting methodology achieved an average True-Positive rating of 0.683 and an overall model accuracy of 74.82% for classifying charts according to levels in Dance Dance Revolution.

Classification of genres is among the important tasks of musical knowledge discovery. It may affect the accuracy of finding results or reducing the processing time when looking for a certain musical genre in an internet context. While the genre classification scheme looks very promising for western genres, the genre of non-western still has no space in genre retrievals, especially in identifying nasheed song. Therefore, a research has been carried out to select the best features to describe nasheed genre and creates a classifier using the selected features to classify nasheed. The features selection technique and the classifier were built based on the theory of fuzzy-soft set that have enough parameters to handle uncertainties in data. The result show that the built classifier using the selected features accurately can identify the nasheed genre up to 90%.

Software-Defined Networking (SDN) has attracted tremendous attention in recent years as the future communication network architecture. However, SDN deployment in legacy network will be progressively phased over a period, especially for larger network which consists of hundred or more nodes. Every migration (i.e. replacing or upgrading) of SDN-enabled nodes requires considerable optimization efforts in terms of cost of investment, network stability and performance gains. Hitherto literatures have proposed variety of static heuristic algorithms to compute the migration sequence of SDN-enabled nodes for multi-periods SDN deployment in legacy network. The aim of each computed migration sequence is aims to improve network performance gains with respect to address different constraints. However, the dynamicity of an unique network, such as traffic growth or topology change, cannot be comprehensively addressed using a static heuristic algorithm over the deployment duration. Machine learning (ML), on the other hand, has been proven successfully applied for various dynamic and non-linear problems in diverse domains. In this article, we summarize the generic workflow for ML in networking domain at first. Subsequently, we investigated the problem of SDN deployment in legacy network from the perspective of ML. We proposed a SDN deployment problem that formulated as Markov Decision Process and reinforcement learning techniques, such as Qlearning and SARSA, can be used to model for the problem.

The usage of LED lighting has been commonly used in indoor and outdoor to facilitate energy-efficient. Because of its characteristics that relatively consume less energy compared to other traditional forms of lighting, it is considered the best alternative to be used as a lighting source. LED lighting also can be applied in stadium lighting applications. The lighting requirement for a stadium is exceptionally high. It requires many luminaires to be used to meet these specifications. The specifications consist of several parameters that inter-related that will affect the performance of lighting. Besides, a different type of view angle of a luminaire will give a different visual performance. Due to this challenge, it requires thorough work being done during the design process as well as during the installation process. The luminaires need to be aimed at proper aiming point coordinates to meet the specifications. Hence, a study on the characteristics of LED lighting needs to be done. In this paper, the objective is to model the output from LED lighting and study the effect of tilt angle and view angle of the luminaire on the visual performance of stadium lighting. A computational model of LED luminaire was developed using MATLAB. With the developed model, the effects of different tilt angles and a different beam angle of luminaire were investigated.

The problem of class imbalance in machine learning occurs when there is a relatively big disproportional distribution of classes in the data for classification tasks. In many real-world domains, such as in healthcare, finance, and predictive maintenance, the number of data points of a less important class (usually the negative class) is much higher than the class of greater interest (usually the positive or target class). This affects the ability of many learning algorithms to find good classification models. To address that, many approaches for solving this problem have been proposed, prominently including ensemble methods integrated with sampling-based techniques. However, these methods are still prone to the negative effects of sampling-based techniques that alter class distributions via over-sampling or under-sampling, which can lead to overfitting or discarding useful data, respectively, and thus affect performance. In this paper, we propose a new data preprocessing sampling technique dubbed as (sBal) for ensemble methods for binary classification in the case of imbalanced datasets. Our proposed method first turns the imbalanced dataset into several balanced bins/bags. Then multiple base learners are induced on the balanced bags and finally, the classification results are combined using a specific ensemble rule. We evaluated the performance of our proposed method on 50 imbalanced real-world binary datasets and compared its performance with well-known ensemble methods that utilize data preprocessing techniques namely SMOTEBagging, SMOTEBoost, RUSBoost, and RAMOBoost. The results reveal that the proposed method brings considerable improvement in classification performance relevant to the compared methods. We performed statistical significance analysis using Friedman’s non-parametric statistical test with Bergman post-hoc test. The analysis showed that our method performed significantly better than the majority of the methods across many datasets, suggesting a better preprocessing approach than the ones used in compared methods. We also highlight possible extensions to the method that can improve its effectiveness.

Augmented Reality Game for based learning has been enhanced the learning experienced and developed the knowledge and skills of the user. The project methodology used in this study is the game development life cycle (GDLC). It includes initiation, pre-production, production, testing, beta testing and release. The purpose of this project is to produce video games for children with dyslexia who have visual and auditory learning difficulties related to memory, time management, speed processing, organization, organization and planning. The objective of this product was to develop Augmented Reality games for dyslexia students, second was to develop Reality-based games on reading, spelling and numbers for dyslexia students using the Unity Game Engine, the third was to test the appropriateness of learning based on reading, spelling games and numbers for dyslexic students. The number of user targets used to test this product are dyslexia students, teachers who teach dyslexia students, expert programmers and designer games and evaluators from the eLearning Carnival & Conference (eLCC 2019). The result of this DyslexiAR learning game is that dyslexic students have a better understanding of learning to read, spell and learn numbers.

The most derived method and realistic way to predict the current stock price is via media resource and trusted new. In this paper, we will apply the current classifier text technique (Based LSTM) and pre-trained model from transfer learning to gain more intuition in financial news and precisely predict stock price. Finally, after using the latest pre-trained word embedding and a classification layer. We have achieved the robust success, and the experiment result shows that our method is able to outperform in accuracy than the previous one and have some advantage in the adaptive dataset.

The goal is to show the usefulness of the 4-point half-sweep EGKSOR (4HSEGKSOR) iterative scheme by implementing the half-sweep approximation equation based on the Grünwald-type fractional derivative and implicit finite difference (IFD) method to solve one-dimensional (1D) time-fractional parabolic equations compared to full-sweep Kaudd Successive over-relaxation (FSKSOR) and half-sweep Kaudd Successive over-relaxation (HSKSOR) methods. The formulation and implementation of the 4HSEGKSOR, HSKSOR and FSKSOR methods are also presented. Some numerical tests were carried out to illustrate that the 4HSEGKSOR method is superior to HSKSOR and FSKSOR methods.

Everyday people receive a lot of information through social media and online news portals. To distinguish whether the information is fake or true is a big problem. An algorithm has been developed to distinguish fake news and true news by searching the relevant news from reliable news website based on the news given. This results in the similarity percentage between news and the relevant news. The algorithm has been tested with the dataset collected by Dr. Victoria L. Rubin that consists of 180 true news and 180 fake news from several American and Canadian news websites. The precision of 69.44% has been achieved with the dataset.

Sentiment analysis is an important branch task of text classification and the related system usually is applied to in perception of user emotion and public opinion monitoring. By comparison, the text classification can be applied to more fields than sentiment analysis. In the system architecture, same as text classification, the complete classification system mainly contains data acquisition, data pre-process, feature extraction, classification algorithm and result output. The Web crawler usually be used in first step, the URL Link, hashtags, Non-Chinese text should be removed in second step. In feature extraction, the IG, TF-IDF, Word2vec usually be used. Then, the SVM, Naive Bayes, KNN or Neural network algorithm usually be used in classifier. Furthermore, as a system that can run automatically, the sentiment analysis system should be able to extract significant feature from corpus and make accurately analysis about emotional polarity of text corpus. At present, the system improvement direction of related system focuses on 3 aspects: data acquisition, feature extraction and classifier algorithm.

This paper presents a numerical method of Newton Successive Over-Relaxation (NSOR) iteration with quadrature scheme to approximate the solution of nonlinear Fredholm integral equations. Here, the quadrature scheme is used to derive the approximation equations of nonlinear Fredholm integral equations in order to develop a system of nonlinear equations. NSOR consists of two parts. In the first part, Newton’s method is used to linearize the developed system of nonlinear equations. Then, in the second part, SOR iteration is used to solve the corresponding system of linear equations to get the approximate solution. In order to validate the performance of the proposed method, Newton-Jacobi (NJacobi) and Newton-Gauss–Seidel (NGS) are used as the reference methods to perform the comparative analysis. Also, some numerical examples are presented to illustrate the validity of the NSOR.

Archiving and storing information is a vital aspect that every government strives to optimize while serving citizens. An electronic document system ensures that the flow of information from storage, processing to transmission is as seamless as possible. The Electronic Document Management System (EDMS) is one such system that governments from around the world have been striving to implement. However, a hindrance has been observed in the acceptance of EDMS by government employees, which has been attributed to several factors. This work highlights a systematic literature review of the factors that affect the acceptance of the system. Additionally, a quantitative approach through a questionnaire is used to determine which factors affect the acceptance of EDMS among government employees. A lack of awareness was established as the primary factor affecting the acceptance of EDMS. It was concluded that the government should put more effort into ensuring that more people are aware of any systems that are implemented by the government.

Document management (DM) is fundamentally one of the most effective approaches applied in managing information flow in an organization. In reality, the utilization of an electronic document management system (EDMS) is depicted essentially when documents are used as memory storage for companies and record-keeping portals that document how operations are achieved. Consequently, the study aimed to respond to the recommendation of McLeod et al. (Arch Manuscr 39:66–94, 2011 [5]) to explore the predictors of EDMS implementation to advise on planning the successful use of EDMS programs. The first objective was to develop a new theoretical model based on the unified theory of acceptance and use of technology (UTAUT) to study the user acceptance and adoption of the EDMS among government employees. The second objective was to determine the factors affecting EDMS acceptance and adoption among government employees that should be given priority using the analytic hierarchy process (AHP) methods. The third objective was to determine whether the constructs of UTAUT influence behavioral intention to use EDMS among government employees. The research employed a systematic review that was developed to identify different factors affecting government employees’ acceptance and adoption of EDMS. The critical success factors were generated based on four levels of UTAUT from 15 articles. The study’s results, from the systematic review, yielded 40 factors that influenced the use of EDMS among government employees.

There are a lot of work has been implemented to solve the problem of text classification but There is only few researchers doing Arabic text classification because of the difficulties in text preprocessing. Convolution Neural network and support vector machine is two different algorithm that can be applied on text classification. CNN seems to be good in extracting the feature from input and SVM is good for classify the class. This study is to introduce Hadith text classification using Convolutional Neural Network and Support Vector Machine. In order to get preliminary result, we used BBC news article (English language) and Arabic tweet sentiment (Arabic language) as dataset for CNN with SVM model. There are 4 methods to evaluate the model which are f1-score, precision and recall and accuracy and error rate probability. We evaluate the model using accuracy and loss using different learning rate. The model accuracy and loss for preliminary result of BBC news article (English language) and Arabic tweet sentiment(Arabic language) are 0.857 accuracy, 0.245 loss and 0.884 accuracy, 0.344 loss. This shows that the proposed model has potential for Hadith text classification.

In this paper, a virtual assistant framework called Alice is presented. This virtual assistant is a combination of 3D avatar, face detection, face recognition and face expression recognition with a voice assistant that similar to Amazon’s Alexa. The 3D avatar (Alice) is a female character animated using Unity and the lip is animated to sync with the speech to make it looks like speaking. Besides that, the 3D avatar can display different facial expressions such as happy, sad and upset. Face detection and recognition makes the system aware of the human user’s identity. Whereas, face expression recognition enables the system to detect the facial expression of the human user. Whenever there is a question being asked, the system will use Speech-to-Text system to convert human speech to text and Natural Language Processing to interpret the intent behind the text. Based on the result of interpretation, the system decides which audio file to be used as response. Then, a realistic artificial voice is generated as response to the human user. The system can access database based on user’s identity to retrieve information about that user. This may create a personalized experience for the human user. This framework can be customized for other applications for different fields. For this Alice framework, two applications have been developed namely a question answering chatbot and a customer service agent.

We evaluate the first-order approximation solution piecewise by first-order polynomial collocation with Quadrature scheme on second-type Fredholm integral equations. This discretization derived the formulation to solve the first order piecewise approximation equation in which the linear system was built. The SOR method was described as a linear solver in which its formulation was constructed and applied in this study. In order to obtain the approximation solutions, the combination of SOR iterative method with the first-order piecewise polynomial by collocation with quadrature scheme has shown that performance of SOR method is superior than Jacobi method in terms of number of iterations and time of completion.

Computer vision has gained tremendous attention recently due to what visual data can provide in terms of meaningful information and predictions. Videos, not like other types of data, can carry lots of information about the captured scene. Information such as objects detection, face recognition and action classification can be beneficial to monitoring systems such as traffic monitoring and security systems. Activities recognition, in particular, is a quite significant part of visual data analysis and can provide pragmatic predictions on people’s behaviour. The absence of a well-labelled video dataset makes it more challenging to develop machine learning algorithms for irregular actions recognition. These prediction models can ease the process of monitoring buildings, roads and other common areas monitored by CCTV systems. This paper proposes a method to utilise deep learning in classifying people’s behavior by identifying normal behaviours and classifying any unusual activities and provide a well-trimmed and labelled dataset for abnormal behaviors in CCTV videos.

Medication non-adherence is one of the most significant concerns in managing chronic diseases which has inevitable consequences. While various technologies and research have been developed and carried out to monitor medical adherence for patients, their approaches lack in terms of the assurance of medicine consumption and the cost effectiveness of their solutions. This paper provides a cloud-based medical adherence system that can track patients’ medicine intake based on the physical effects of the medicine on their bodies by tracking their vital signs. A machine learning model is trained to classify the patient health status and this data is used to determine whether their bodies are responding to the medicine, which is used to alert doctors to enable home hospitalization. The use of this system is proposed to serve as a secondary decision support provider to compliment and ease the decision-making process done by doctors.

Biometrics is an effective way to identify and authenticate users based on their personal traits. Among all kinds of hand-based biometrics, finger vein appears to be emerging biometrics that has received a great attention due to its rich information available and ease for implementation. With finger vein system becoming more and more popular, there have been various attempts to comprise the system. Recent studies reveal the vulnerabilities of finger vein system to presentation attack where the sensory device accepts a fake printed finger vein image and gives access as if it were a genuine attempt. In this study, a presentation attack detection method based on hybrid feature spaces of finger vein texture analysis is proposed. Histogram of oriented gradient operator is applied on different channels of grayscale and color feature spaces to obtain texture information of the histogram descriptors. The proposed method includes two implementations of feature space analysis, namely CHOG1 and CHOG2. A well-established publicly available dataset is used to analysis and evaluate the proposed implementations. Experimental results suggest that the combination channels of grayscale and color luminance is able to generate better performance through Support Vector Machine classifier with ACER as low as 0.60% and 0.74% for CHOG1 and CHOG2, respectively. The experiments show that the implementation of CHOG1 performs slightly better than single channel max gradients of CHOG2.

Since an ever-increasing part of the population makes use of social media in their day-to-day lives, social media data has been analyzed in many different disciplines. While there is a great deal of literature on the challenges and difficulties involving specific data analysis methods, there hardly exists research on analyzing the appropriate techniques used to handle different types of data for the purpose of social media analytics. To address this gap, we conducted an extended and structured literature analysis through which we identified challenges addressed and solutions proposed. The literature search revealed that three types of data that were least used for social media analytics that includes Bluetooth, WIFI and mobile roaming data. In contrast, other types of data have received more attention. Based on the results of the literature search, we discuss the most important challenges for researchers and present potential solutions. The findings are used to extend an existing framework on social media analytics. The article provides benefits for researchers and practitioners who wish to collect and analysis social media data.

Real-life stressors such as work pressure and examination exist in daily life and will affect heart rate (HR) and heart rate variability (HRV) of an individual. The objective of this study is to evaluate the effect of stress on HRV values in healthy human subjects. Wearable device equipped with photoplethysmography (PPG) sensor is worn by all subjects to record the HR data in two situations, which are during rest situation and stress situation for a period of 10 min in each situation. The recorded HR data were then analysed using MATLAB and Kubios HRV software to get the HRV values. Results obtained show that all subjects have higher HRV values during rest situation and those values drop drastically when subjects were exposed to stress simulation. HRV calculated from the root mean square of successive differences (RMSSD) values are more stable and consistent in determining the HRV values compared to standard deviation of the nearest neighbor intervals (SDNN) values. Besides, all subjects show changes in interbeat interval (IBI) from high fluctuations to low fluctuations when subjects are in stress situation. Low fluctuations in IBI changes will result in lower HRV values indicating the presence of stress components. In conclusion, this study provides evidence that one will have lower HRV values during stress situation compared to rest situation.

The linear rational finite difference method (LRFD) is becoming more and more popular recently due to its excellent stability properties and convergence rate, especially when we are approximating the derivative of some points near the end of the interval. The main intention of this paper is to combine the 3-point linear rational finite difference (3LRFD) method with the composite trapezoidal (CT) quadrature formula to discretize the first-order linear integro-differential equation and produce dense linear systems. Furthermore, the numerical solution of the integro-differential equation is obtained by implementing the Successive Over-Relaxation (SOR) method. At the same time, the classical Gauss–Seidel (GS) method is also introduced as the control condition. In the end, through several numerical examples, the number of iterations, the execution time and the maximum absolute error are compared, which fully illustrated the superiority of SOR method over GS method in solving large dense linear system generated by the CT-3LRFD formula.

In this article, we propose semi-approximate approach in finding a solution of Burgers’ equation which is one of the partial differential equations (PDEs). Without using the Newton method for linearization, we derive the approximation equation of the proposed problem by using second-order implicit scheme together with the semi-approximate approach. Then this approximation equation leads a huge scale and sparse linear system. Having this linear system, the Successive Overrelaxation (SOR) iteration will be performed as a linear solver. The formulation and execution of SOR iteration are included in this paper. This paper proposed four examples of Burgers’ equations to determine the performance of the suggested method. The test results discovered that the SOR iteration is more effective than GS iteration with less time of execution and minimum iteration numbers.

In this paper, the Redlich-Kister (RK) polynomial interpolation have been formulated and analyzed in solving two-point boundary value problems (BVPs). The Redlich-Kister polynomial interpolation is tested with certain number of different sizes and compared with Cubic Trigonometry B-Spline Interpolation Method (CTBIM) and Power Polynomial (Power). To do that, the discretization process of BVPs by imposing the generated RK dense linear system. Then this dense linear system need to be solved via direct method to determine the approximate value of unknown coefficients in which these coefficient are used to form the RK approximation function. Based on the maximum norm (MaxNorm) and L2-Norm, the results showed that the solution by using the RK approximate function is the more accurate compared with CTBIM and Power methods.

Undoubtedly, the initiative of the Malaysia education ministry to introduce coding in school curricula is a very good effort with lots of advantages. Able to code will be an advantage and a necessity when the students join the workforce. After not small amount of budget and time has been spent by the country for this mission, there are several issues that need to be considered and worked on to ensure coding lessons in schools achieves the target. This paper presents six issues and the recommended solutions that do not require a change to the current educations system as frequent changes in national education policy will burden teachers, parents and students. The identified issues are from the perspective of language of communication, implementation and execution, digital divide, quality tools, assessments and teaching and learning time. This works suggests to create a self-learning system built specifically for the national secondary schools’ syllabus, short-term job exemption for teachers and programming skill test to replace project as part of assessments to increase the rate of teaching effectiveness of Computer Science subjects, especially programming. This study will be of great important to educational planners, school authorities, educational researchers and the governments.

This paper purposely attempts to solve two-dimensional (2D) parabolic partial differential equations (PDEs) using iterative numerical technique. Also, we determine the capability of proposed iterative technique known as Successive Over-Relaxation (SOR) iteration compared to Gauss–Seidel (GS) iteration for solving the 2D parabolic PDEs problem. Firstly, we transform the 2D parabolic PDEs into 2D elliptic PDEs then discretize it using the similarity finite difference (SFD) scheme in order to construct a SFD approximation equation. Then, the SFD approximation equation yields a large-scale and sparse linear system. Next, the linear system is solved by using the proposed iterative numerical technique as described before. Furthermore, the formulation and implementation of SOR iteration are also included. In addition to that, three numerical experiments were carried out to verify the performance of the SOR iteration. Finally, the findings show that the SOR iteration performs better than the GS iteration with less iteration number and computational time.

This paper discusses the implementation of “JKalvi”, an E-Learning approach used to replace traditional teaching and learning methods through the use of E-Games. Although the use of educational technology in education is leveraging, there are still gaps in the use and implementation of such approaches being utilized for teaching and learning especially in high school education. In this study, users, both educators and students can utilize the resources in the application in increasing the interactivity of learning through the use of online materials and games all in a simple android platform. Given the increased use of android based mobile phones among high school students today, users can monitor signs of progress and performance through a digitized system replacing traditional paper-based methods all from a click away through their mobile phone. This is persistent with the examination of students’ perceptions in the pursuit of new learning methodology through the use of this application conducted through via surveys. A total of 30 preliminary surveys were recorded and analyzed using descriptive statistics. The findings indicate that 90% of the students agreed that the application was interactive, added value to their learning process and found no issues in the use of the application.

The bee honey industry is a very lucrative industry and in just 2015, Malaysia imported RM58 millions ringgit worth of honey products. Although stingless bee keeping seemingly simple, there are a few problems that could affect the failure of success. Stingless bee industry in Malaysia are facing major challenges especially in the respect of the queen, quality inconsistency, low honey production and high price. This is because most of the stingless beehive farm in Malaysia do not have a proper system to monitor and control the environmental parameter of beehive. Moreover, because of the medicinal value of the stingless beehive, there are also a few cases of stolen beehive in Malaysia. Therefore, to prevent these issues, a project named Smart Stingless Beehive Monitoring System (SSBMS) is being developed to monitor and control the environmental parameter of beehive automatically. Sensors are installed at stingless beehive to monitor and control the environmental parameter of stingless beehive such as temperature, humidity, water level as well as the geographical location of beehive through Internet of Things (IoT) platform. The feedback system such as DC fan motor and diaphragm pump have been implemented to ensure the temperature of stingless beehive and water supply for stingless bee can be maintained below 30 °C and in an adequate level respectively. In conclusion, SSBMS is able to help reducing manpower in monitoring the environmental parameter of stingless beehive, gives feedback to maintain the internal beehive temperature and water supply level automatically or based on beekeepers’ desires and tracks the geographical location of the stingless beehive through Global Positioning System (GPS).

Improving machine learning algorithms has been the interest of data scientists and researchers for the past few years. Among the performance problems raised is the classification imbalance issues listed as the top ten. The present study makes comparison and analysis of 5 state-of-art classifiers, 5 ensembles classifiers and 10 resampling techniques for data imbalance. This is done via the used 4413 instances consisting of demographic, economic, and behavioural data from student information systems and e-learning, as well as engineering faculty for the first semester 2017/2018. The use of three sampling types was adapted for the analysis: oversampling, undersampling and hybrid. The experimental results prove to model students’ behaviour from e-learning data and bagging decision tree ensemble classifier produces the highest results. Lastly, a hybrid resampling technique, SMOTEENN consistently shows the top result compared to other resampling techniques.

The amalgamation of atmospheric elements indicates positive trends in sea level rise which has had a significant impact on nearly 60% of the world’s population living in the low elevated coastal area. In this paper, we first discuss potential factors leading to the rise in sea level and negative impacts on future development along the coastal region. Then, methods of acquiring sea level data which revolutionize the study of variation at sea level will also be reviewed and discussed. The present paper aims to review several Deep Learning (DL) algorithms that address critical issues of forecasting, specifically a time variable known as time series by managing complex patterns and inefficiently capturing long-term multivariate data dependency. Asynchronous data handling required correct theoretical framework processes. Based on the review conducted, the deep learning architecture is capable of generating accurate prediction at sea level which can be used as decision-making tools for managing low-lying coastal areas.

Education is one of the important fields in this challenging world. The researchers come out with the new perceptive, which is learning analytics that is a new invention for helping out the instructors, learners, and administrators. The use of learning analytics can be the medium for increasing the productivity of education for producing capable leaders in the future. Machine learning comes out with any type of techniques such as Decision Tree, Support Vector Machine, Naïve Bayes, and Ensemble Classifiers. However, both Decision Tree and Ensemble Classifiers are chosen as the best potential machine learning techniques to cope with the large database of students. The Boosted Tree of Ensemble Classifiers managed to get 99.6% accuracy of training 378,005 data of students regarding the Virtual Learning Environment (VLE).