Proceedings of the 7th International Conference on Communications and Cyber Physical Engineering

Table of Contents

1. Frontmatter

2. Early Enhancements Within IoT Technology Integration for Ecological, Security, and Economic Auditing in Underground Mines: A Review

   J. Chinna Babu, Adarsh Kumar, J. Mounika, T. Naresh Kumar, Y. Sudheer Kumar, S. V. V. Prasad

   Abstract

   The existence of dangerous materials such as coal ash and greenhouse gases, responsible for countless mining-related tragic events, makes underground coal mining safety a top priority. This study emphasizes the necessity for real-time environmental monitoring and safety measures. It demonstrates wireless sensors in conjunction with Internet of Things (IoT) technologies to develop a comprehensive framework for monitoring mining operations. The emphasis is on cost-effective solutions that make use of sophisticated technology such as Arduino-based sensors, Zigbee networking, and RFID tags. It intends to modernize industry safety standards by applying this revolutionary technology and methodology, solving long-standing concerns about gas dangers, environmental conditions, and accident prevention. This comprehensive technology is a critical invention that can completely change the underground coal mining industry by giving employees a safer and more productive work environment and meeting the world’s growing demand for coal production.

3. LoanPredictor: A Streamlit Application for Interpretable and Fair Loan Approval Predictions

   T. N. Ranganadham, SaiKrishna Ulavapati, Mg Pandu Ranga Reddy, SaiVinod Anandi

   Abstract

   LoanPredictor is an innovative Streamlit-based application designed to predict loan approval outcomes with a strong emphasis on interpretability and algorithmic fairness. Leveraging advanced machine learning models, the system provides transparent explanations for its decisions, enabling users and stakeholders to clearly understand the key factors influencing loan approvals. The application integrates fairness-aware algorithms to actively minimize biases, ensuring equitable treatment across sensitive attributes such as income, marital status, and gender. Through a streamlined, user-friendly interface, individuals can input their personal and financial data to receive accurate and comprehensible predictions. Built on Streamlit, the platform supports intuitive user interaction and rapid prototyping. By embedding ethical AI principles into the loan evaluation process, LoanPredictor empowers financial institutions to adopt transparent, fair, and responsible decision-making practices—ultimately transforming traditional lending into a more inclusive and accountable system.

4. Robust Fusion of Multimodality Medical Images with Fidelity Driven Optimization and Detail -Preservation

   M. Ravi Kishore, K. Shankar, K. Rajmohan, Y. Pavan Kumar Reddy, Y. Sunanda

   Abstract

   Medical image fusion is an important computer-aided technique that combines critical elements from diverse medical pictures to improve diagnostic detail and accuracy. Despite recent advances, many fusion systems fail to provide enough information and textures for accurate disease identification, generally due to insufficient noise removal from the original medical image. In response to these problems, we offer a unique fusion technique designed for multimodal medical imaging. Our method uses fidelity-driven optimized (FDO) reconstruction to maintain important information while reducing noise impacts. Furthermore, we present a rank-coefficient optimization strategy for reducing the impact of noise on multiple mediums medical imaging Furthermore, we recommend the use of a repetitive detail preservation strategy to incorporate extra data as well as textures in the initial multimodal medical images while retaining high resilience. The final product demonstrates the capabilities of our innovative fusion technology. Extensive studies have proved the resilience of our technique, notably in dealing with noisy medical images, highlighting its potential for use in diagnostic applications.

5. Implementing Supervised Machine Learning Algorithms on an Investigation of Customer Mobility in the Telecom Sector

   Anuradha Boya, Maddela Parameswar, R. Suhasini, Modugula Siva Jyothi, D. T. V. Dharmajee Rao, B. Thanuja

   Abstract

   The loss of customers, commonly referred to as attrition, is one among the most significant issues the telecoms industry has to deal with. Getting new customer costs more than maintaining an old one. Data analysis can be used to identify customer churn causes and retain customers by collecting data from telecom companies. Therefore, for telecom firms to retain their consumers, churn forecasting is essential. In order to keep their current clients, telecom providers need to be aware of what generates churn. The knowledge needed to gain this information can be extracted from telecom data. We will develop and analyze a number of machine learning models that are supervised in this, and we can say that Random Forest performs the worst compared to others because they offer greater accuracy.

6. Development and Analysis of a Better AODV Routing Protocol for Improved Network Performance in Manets Using a Machine Learning Model

   V. Narasimha, C.H. Rajakishore Babu, Golla Saidulu, K. Srujan Raju, Mohan Babu Bukya, Kotha Chandrakala

   Abstract

   A MANET is a self-configurable wireless ad hoc network that does not support demobilization. The route is susceptible to many security vulnerabilities because of its adaptability. IDS performance should be changed as a result to remedy this. In this study, a method for increasing IDS’s precision and rate of identification for the AODV routing protocol is presented. It relies on the Machine Learning (ML) method. In contrast to previous assaults, wormhole assaults use a secret, out-of-band channel to initiate the attack, making detection more difficult. This kind of assault does not involve any cryptographic weaknesses on the part of the attacker. One type of Denial of Service assault is the wormhole assault.

7. A Review on Early Prediction of Heart Failure Using Machine Learning Models

   Sherin Anaya, Ananya Mazumder, V. Sreemedhini, S. Srinikha, P. Iyappan

   Abstract

   Many individuals suffer from heart failure, which is a recurrent issue worldwide. Early detection of this illness is crucial in determining the best course of action. Advising early prediction of heart failure can help prevent life-threatening situations. In recent times, machine learning algorithms have been utilized to forecast diverse types of ailments. Machine learning models are trained using datasets and tested using parameters that have been medically verified. Numerous researchers have put forth various approaches utilizing machine learning (ML) for early detection based on a few metrics, such as age, gender, blood pressure, BMI, diabetes, Hb level, and many more. The goal of our study is to compare various existing ML models considering the categories like Accuracy, F1 Score, Recall and Precision. To predict the early detection of heart failure, the following ML Algorithms are considered, evaluated, and compared in the study: Random Forest Classifier (RFC), Naive Bayes, XGBoost, K Nearest Neighbour (KNN), and Logistic Regression (LR). These ML algorithms are implemented in various data sources, and their performance is closely monitored. According to the findings, Random Forest Classifier (RFC) is preferred for prediction.

8. A Peer Review on MPPT Extraction in Structured Hybrid Power Systems with Optimization Topologies

   K. Rajesh Kumar, R. Sripriya, S. K. Bikshapathy

   Abstract

   The escalating concern of global warming, primarily caused by an excessive presence of carbon dioxide (CO2) in the Earth’s atmosphere, along with the limited accessibility of fossil and nuclear fuels on a worldwide level, has intensified the necessity to explore alternative energy sources in order to fulfil forthcoming energy requirements. The investigation of alternative energy sources is of utmost importance in mitigating adverse environmental impacts and addressing the continuously growing need for electricity. The demand for renewable energy sources arises from the convergence of environmentally sustainable power generation and the economic considerations associated with non-renewable energy. The article offers a comprehensive analysis of the research conducted on the modelling, optimization, and control aspects of a self-contained hybrid power system. It also provides an extensive review of the various methods utilized for maximum power point tracking hybrid systems

9. Crop Recommendation System Using Soil Content and Weather

   P. Phanindra kumar Reddy, V. Madhu Sudhan Reddy, K. Nithin, P. Visweswara Rao, K. S. Maajid Hussain

   Abstract

   This research analyzes the soil properties in detail by employing the techniques of data science, taking into account variables like temperature, moisture content, and chemical composition. Its main goals are to assess crop quality, identify weed infestations in a variety of plant species. The first step of the process involves exploratory data analysis, where missing values are identified using tools such as HeatMap. Using a range of machine learning models, such as Support Vector Classifier, Random Forest, KNN, and decision trees, the system uses random forest to achieve high accuracy (99%) in crop recommendation. In order to anticipate future 5-day data, it also integrates weather forecasting skills. In addition to offering a comprehensive approach to maximize agricultural yields, the project guarantees user-friendliness by integrating an easy-to- use interface for smooth communication with the system.

10. Enhancing Phishing Detection with Optimized Data Workflows and Machine Learning

    Shubham Kean, Durga Prasad Banoth, Akhil Sapavath, Subhani Shaik

    Abstract

    As phishing attacks become increasingly sophisticated, mimicking legitimate communications with alarming precision, the need for advanced cybersecurity measures has never been more urgent. Traditional defense mechanisms, often constrained by their static, rule-based nature, fall short of the dynamic and adaptive strategies employed by cyber attackers. This study introduces an innovative approach to phishing detection that leverages the power of machine learning (ML) alongside optimized data workflows, marking a significant leap in the capability to identify and counteract phishing attempts with heightened accuracy and markedly fewer false positives. Through the development of a specialized data pipeline designed for ML-driven detection and the strategic application of feature engineering across diverse datasets, we significantly bolster the system’s adaptability and scalability. This novel integration not only enhances the precision of phishing detection but also imbues the system with the agility needed to respond to emerging threats promptly. The results of our research not only demonstrate a robust advancement in phishing detection capabilities but also lay the groundwork for future innovations in adaptive, data-centric cybersecurity defenses.

11. Performance Analysis of Electric Vehicle Battery Using Artificial Neural Networks

    Neelima Kalahasthi, Sri Chandana Kanthi, Nandini Guntipally, Rohith Reddy Bongarapu, GopiKrishna Kopati, Gangapuram Saikumar

    Abstract

    The electrification of transportation has spurred significant interest in improving the performance and efficiency of electric vehicle (EV) batteries. Predictive modelling techniques, such as Artificial Neural Networks (ANN), offer a promising approach to analyze and predict EV battery performance. This study presents a comprehensive analysis of EV battery performance using ANN, focusing on key performance metrics such as capacity, charging time, and efficiency. The study utilizes a dataset comprising historical data collected from EV batteries, including factors such as temperature, charging rate, and battery age. The ANN model is trained using this dataset to predict battery performance in various settings. The outcome demonstrates Considering the ANN model can accurately predict battery performance metrics, with high levels of accuracy and reliability. Furthermore, the study investigates the impact of various factors on battery performance. It is found that temperature has a significant influence on battery capacity and charging time, with higher temperatures leading to reduced capacity and longer charging times. Similarly, charging rate is found to impact battery efficiency, with higher charging rates leading to lower efficiency. The study also explores the potential use of the ANN model for optimizing battery performance. By analysing the ANN predictions, it is possible to identify optimal operating conditions for EV batteries, such as temperature and charging rate, to maximize performance and efficiency. Overall, this study highlights the effectiveness of using ANN for predictive analysis of EV battery performance. The results demonstrate the potential of ANN as a valuable tool for battery manufacturers and EV manufacturers to optimize battery design and operation. By leveraging ANN models, manufacturers can improve battery performance, increase efficiency, and enhance the overall driving experience for EV owners.

12. Advancements in Lung Cancer Prediction: An In-Depth Analysis and Optimization of Artificial Neural Network Algorithms

    B. Pandu Ranga Raju, J. Hemalatha, S. Sai Keerthana, K. Madhavi, B. Himaja

    Abstract

    Using cutting-edge neural network designs, the study “Advancements in Lung Cancer Prediction: An In-depth Analysis and Optimization of Neural Network Algorithms VGG16, InceptionV3, and EfficientNetB0” explores the field of lung cancer prediction. This study examines the performance of three well-known models: VGG16, InceptionV3, and EfficientNetB0, with an emphasis on improving prediction accuracy. By means of comprehensive research and optimization methodologies, the initiative endeavors to attain noteworthy advancements in precision levels. The experimental findings demonstrate the usefulness of these algorithms in lung cancer prediction tasks, with obtained accuracies of 80%, 77%, and an astonishing 93%, respectively, indicating promising results. With the potential to improve patient outcomes and healthcare practices, this research adds to the continuing efforts to use machine learning for early identification and prognosis of lung cancer.

13. Diagnosis of Fetal Brain Abnormalities Using Ex Learning

    Ch. Sravanthi, K. Kavya Sri, V. Varshitha, V. Keerthi

    Abstract

    This work aims to develop a model using Convolutional Neural Network (CNN) for classifying fetal ultrasound images to diagnose fetal brain abnormalities. Ultrasound can identify the majority of major structural fetal abnormalities. By using TensorFlow and Keras frameworks, it preprocess the dataset and construct a CNN architecture which has convolutional and fully connected layers, and optimizes the model’s accuracy with exceptional precision. Through continuous training and validation, this model gains the ability to correctly identify foetal pictures as normal or abnormal. This helps in early detection of anomalies and intervention for expectant mothers, thereby improving outcomes for both mothers and babies.

14. Mediguard: Blockchain-Enhanced Security for the Internet of Medical Things

    N. Swathi, A. Harika, G. Gouthami, K. Amrutha Sree, D. Swetha

    Abstract

    The functionality of smart healthcare systems has been greatly improved by the quickly developing Internet of Medical Things (IoMT), with its cutting-edge real-time services. However, there are significant privacy and security problems with IoMT, which are exacerbated by the wide range of these devices, making it more challenging to provide a unified security solution. Moreover, the prevalent cloud-centric IoMT healthcare systems, which rely on cloud computing for Electronic Health Records (EHR) and medical services, make a decentralized IoMT healthcare model impractical. This study proposes an innovative blockchain-based architecture that ensures decentralized EHR and smart contract-driven service automation while preserving system security and privacy. Cloud-centric IoMT healthcare systems suffer from excessive latency, storage costs, and single point of failure. These issues are resolved by combining a blockchain-based distributed data storage system with the recently suggested hybrid computing paradigm.

15. Cloud Secrecy: Techniques for Improving Data Security and Protection in Cloud Capacity

    T. N. Ranganadham, J. Mahendra Reddy, L. Kiran, B. ManiKanth, B. Lokeswar, R. Madhan Mohan

    Abstract

    The blast of information that has been delivered as a result of technologies such as the Web of Things (IoT), savvy cities, and digital changes has driven a rise within the capacity industry that has seen huge extensions. The Web of Things (IoT), smart cities, and advanced changes have all contributed to an increase in the number of modern manifestations, which in turn has quickened the expansion of the capacity of commerce. An expanding number of individuals, enterprises, and governments are moving their information to the cloud in order to make strides in productivity and openness. As a result, cloud capacity frameworks have become a fundamental component of the cutting-edge world. There are many dangers associated with this process, including the possibility of information leakage, invasion of privacy, and unauthorized access. The purpose of this theory is to collect different information regarding security measures, protection and access.

16. Predicting Stock Prices with Advanced Deep Learning Technique: An LSTM-Based Approach

    A. Ramesh Babu, G. Shanmukha Sarma, G. Kiran Kumar, A. Madhu, S. Nayab Rasool

    Abstract

    Market forecasting is a multifaceted endeavor influenced by a myriad of factors, blending rational analyses with the ebbs and flows of capitalist sentiment and market dynamics. In this paper, we delve into the realm of data analytics, specifically exploring the efficacy of Long Short-Term Memory (LSTM) deep learning models in accurately predicting stock market prices. Recognizing historical data as a pivotal element shaping market events, we employ machine learning techniques, particularly LSTM, to discern patterns and glean insightful predictions. Our proposed model endeavors to refine investment decision-making by furnishing forecasts of future stock prices with enhanced precision. Through rigorous experimentation and comparative analyses, we demonstrate the superior performance of LSTM over conventional models such as Linear Regression, ARIMA, and Averages, culminating in more robust and reliable predictions.

17. Enhanced Traffic Sign Recognition and Road Lane Detection with Semantic Segmentation for Autonomous Cars

    M. Lavanya, S. Varalakshmi, Venkata Krishna Sravya V. K. Ramachandrula, D. Shailender, A. Shiva Sai Rao

    Abstract

    Autonomous vehicle navigation demands accurate perception systems capable of viewing the real-time scenes in the surroundings. In our research, we present an integrated strategy to semantic segmentation, incorporating traffic sign recognition and lane detection, vital for safe and efficient autonomous driving. Using deep learning techniques, our model employs a multi-class segmentation architecture trained on diverse datasets containing various environmental conditions. Lane detection is done through a combination of computer vision methods and deep learning algorithms, ensuring robust performance across different road types and lighting conditions. Similarly, traffic sign recognition utilizes a fusion of template matching and deep learning-based classifiers. SVM classifiers are trained for each task and integrated into the autonomous vehicle system, creating a real-time processing pipeline. Classification to accurately identify and interpret traffic signs in the scene. Through extensive testing and evaluation, we showcase the effectiveness and reliability of our integrated system in real-world driving environments, achieving high accuracy in tasks such as semantic segmentation, lane detection, and traffic sign recognition. Our systematic presentation represents a significant advanced perception systems essential for safe and autonomous navigation of vehicles in diverse road environments.

18. Pick Fresh-Mobile App for Street Vendors

    Ravi Prakash Reddy, A. Sanjana, G. Sanjana, K. Sravya

    Abstract

    The Pick Fresh application serves as a versatile platform, enabling users to discover nearby sellers and submit product requests seamlessly. Supporting multiple languages, including Tamil, Telugu, Kannada, Hindi, and English, the app empowers sellers to create profiles and add items to their lists. Users can efficiently browse local sellers and their offerings. Upon submitting a product request, sellers can choose to accept it, triggering the redirection of their map location to the customer's vicinity. Users receive notifications when sellers are within a 1-km radius. This innovative application enhances localized commerce by bridging language barriers, connecting users with nearby sellers, and facilitating effective communication. The integration of location-based notifications, multilingual support, and streamlined request handling contributes to an enhanced user experience, fostering efficient interaction between buyers and sellers.