Authors: Diya Manoj, Dr. Asha K

Abstract: Electrocardiography (ECG) is an essential tool for diagnosing heart conditions, yet traditional manual interpretation takes more time and is subject to variability among experts. This research explores the application of deep learning techniques to automate ECG classification, aiming to enhance diagnostic correctness and speed and reliability. Using a dataset comprising 120,000 ECG images, a deep learning model based on the ResNet18 architecture was developed to categorize ECG signals into four classes: Myocardial Infarction, Abnormal Heartbeat, History of Myocardial Infarction (MI), and Normal. The study involved extensive pre-processing of ECG images, including normalization, augmentation, and noise reduction techniques to improve data quality. An exploratory data analysis (EDA) phase was conducted to visualize class distributions and identify potential challenges such as class imbalance. The model was trained for 40 epochs, achieving a training correctness of 99.85% and a best test correctness of 76.85%. Evaluation metrics such as precision, recall, and F1-score were used to assess performance, with confusion matrices revealing areas of improvement. Despite promising results, challenges such as class imbalances, overfitting, and the difficulty of distinguishing similar ECG patterns were encountered. Strategies such as weighted loss functions, dropout layers, and hyperparameter tuning were employed to mitigate these issues. The study concludes that deep learning models can serve as effective tools for ECG classification, providing a foundation for real-time clinical applications. Future work will focus on dataset expansion, model generalization, and real- time deployment to facilitate broader adoption in healthcare settings.

 

 

Published by: vikaspatanker

Authors: Mr. Tanmay Jaiswal, Ashish Koli, Dr. Jasbir Kaur, Ms. Ifrah Kampoo

Abstract: Convenience has emerged as the main factor influencing the adoption of technology in today's fast-paced environment. With the help of a straightforward mobile interface, customers can now order food from a number of restaurants thanks to the substantial evolution of the food delivery app market. The creation of a meal delivery app with user-friendly features like different payment methods, real-time tracking, and a smooth user interface is covered in this article. This application seeks to address the growing need for efficient and dependable meal delivery services by leveraging contemporary mobile development technology.

 

 

Published by: vikaspatanker

Authors: R.S. Deshpande, Kirti Gautam Latake, Santosh Gaikwad, Arshiya Khan,

 

 

Abstract: Sentiment analysis has emerged as a vital technique in natural language processing, enabling machines to identify and interpret emotional tone in text. This paper presents a comprehensive review of traditional machine learning and recent deep learning approaches for sentiment classification. It highlights the evolution from classical algorithms to advanced models like LSTM and ALBERT, emphasizing their accuracy and applicability. An experimental evaluation using the Amazon Fine Food Reviews dataset confirms the superiority of deep learning models. The paper also discusses key challenges and outlines future directions to enhance the interpretability, scalability, and ethical use of sentiment analysis across various domains.

DOI: http://doi.org/

 

 

Published by: vikaspatanker

Authors: Mrs. R. Bhuvaneswari., Ms. K.Lavanya

 

 

Abstract: As network infrastructures continue to expand, the complexity and frequency of cyber threats have significantly increased, highlighting the need for more effective Intrusion Detection Systems (IDS). This study introduces a hybrid approach combining an Enhanced Convolutional Neural Network (CNN) with Linear Regression to identify and categorize network intrusions such as BENIGN traffic, DoS Slowloris, and DoS Hulk attacks. Unlike conventional IDS frameworks that often suffer from high false alert rates and inadequate feature processing, the proposed model utilizes deep learning to extract meaningful spatial features from traffic data. The CNN component captures intricate patterns, while Linear Regression aids in refining classification by pinpointing key behavioral indicators of attacks. Evaluations show that this approach delivers improved detection accuracy, faster anomaly identification, and fewer false positives. Its real-time performance and flexibility make it well-suited for use in cloud-based platforms, enterprise systems, and IoT-driven environments.

DOI: http://doi.org/

 

 

Published by: vikaspatanker

Authors: Rohith Bommalla Naresh

Abstract: Accreditation in higher education faces challenges in adapt- ing to technological advancements and evolving learner expectations. This article explores how artificial intelligence (AI), blockchain, and micro-credentials can transform quality assurance into a dynamic, trans- parent, and inclusive system. Drawing on recent research, global case studies, and theoretical insights, we propose a forward-looking ac- creditation model leveraging AI for real-time evaluation, blockchain for secure credentialing, and micro-credentials for flexible learning path- ways. Despite their potential, these technologies raise ethical, equity, and adoption challenges. A novel framework is presented, emphasiz- ing continuous improvement, global interoperability, and accessibility. Research findings highlight the impact of these innovations, ensuring higher education aligns with a rapidly changing global landscape.

 

 

Published by: vikaspatanker

Authors: Rohith Bommalla Naresh

Abstract: Accreditation in higher education faces challenges in adapt- ing to technological advancements and evolving learner expectations. This article explores how artificial intelligence (AI), blockchain, and micro-credentials can transform quality assurance into a dynamic, trans- parent, and inclusive system. Drawing on recent research, global case studies, and theoretical insights, we propose a forward-looking ac- creditation model leveraging AI for real-time evaluation, blockchain for secure credentialing, and micro-credentials for flexible learning path- ways. Despite their potential, these technologies raise ethical, equity, and adoption challenges. A novel framework is presented, emphasiz- ing continuous improvement, global interoperability, and accessibility. Research findings highlight the impact of these innovations, ensuring higher education aligns with a rapidly changing global landscape.

 

 

Published by: vikaspatanker

Authors: Chirayu C.Jadhav, Omkar K. Pol, Rushikesh M. Amane, Associate Professor Mrs. M. M. Raste

Abstract: The rapid advancement of deep learning has enabled the creation of hyper-realistic synthetic media, commonly known as deepfakes, which threaten digital trust, privacy, and security. While these technologies demonstrate the potential of generative models like GANs, their misuse for misinformation and identity fraud necessitates robust detection methods. This paper presents a comprehensive analysis of state-of-the-art deepfake generation techniques and their countermeasures, focusing on the challenges of distinguishing manipulated content from authentic media. We evaluate data-driven detection approaches, including artifact-based analysis and deep neural networks, highlighting their strengths and limitations under varying compression levels and dataset scales. Building on existing benchmarks like FaceForensics++ and Celeb-DF, we propose a systematic framework for assessing detector performance, emphasizing the role of domain-specific features (e.g., facial micro-expressions, inconsistent lighting) in improving accuracy. Our experiments demonstrate that hybrid methods—combining spatial-temporal analysis with adversarial training—outperform human observers and single-modality detectors, particularly in cross-dataset scenarios. Finally, we discuss emerging threats, such as adaptive deepfakes designed to evade detection, and outline future directions for scalable, real-time solutions. This work aims to standardize evaluation metrics and inspire novel research to safeguard digital media integrity in an era of escalating synthetic threats.

 

 

Published by: vikaspatanker

Authors: Tejitha Pukkalla, Professor Dr. M. Sumender Roy

Abstract: Classifying sleep disorders is crucial for improving individuals' quality of life. Apnoea and sleep disturbances can have a profound effect on a person's health. The classification of sleep stages by experts in the field is a meticulous task that is susceptible to human error. Developing accurate algorithms for machine learning applications (MLAs) aimed at classifying sleep disorders requires thorough analysis, monitoring, and diagnosis of these disorders. To categorize sleep disorders, this research compares traditional MLAs with deep learning algorithms. This study proposes an effective method for classifying sleep disorders, utilizing the Sleep Health and Lifestyle Dataset, which is available online for evaluating the proposed model. The optimizations were performed by adjusting the parameters of various machine learning algorithms using a genetic algorithm. An assessment and evaluation of the proposed algorithm's classification performance were conducted against state-of-the-art machine learning techniques for sleep disturbances. The dataset comprises 13 columns and 400 rows containing various sleep-related variables. Additionally, routine tasks were analysed. The random forest, decision tree, support vector machine, k-nearest neighbours, and deep learning algorithms employing artificial neural networks (ANNs) were assessed. The results of the experiment reveal significant differences in the performance of the algorithms examined. The proposed algorithms achieved classification accuracies of 83.19%, 92.04%, 88.50%, 91.15%, and 92.92%, respectively. The ANN excelled in precision, recall, and F1-score metrics, achieving the highest classification accuracy of 92.92%. The corresponding values for precision, recall, and F1-score were 92.01%, 93.80%, and 91.93%. The ANN algorithm demonstrated superior accuracy compared to other tested algorithms.

 

 

Published by: vikaspatanker

Authors: Shubham Ballal, P.R Sonawane, Utkarsh,, Aashutosh Rawat, Deewan Singh

 

 

Abstract: In modern world, images are among the most significant sources of shared information.They include important information that even those who are illiterate can understand. The growing availability of advanced image editing tools has made detecting image forgeries a crucial problem in digital forensics.However, the majority of Forgery detection methods are limited to identifying a single kind of forgery, like image splicing or copy-move, which are not used in everyday life. In order to improve digital image forgery detection, this paper suggests a deep learning technique that combines CNN and ELA to simultaneously detect two types of image forgeries. The suggested method depends on determining the forged area’s com- pressed quality, which typically varies from the image’s overall compressed quality.The matrix subtraction of the original image compressed image is used as input to CNN model for training and detection. This research paper fine-tunes the CNN and uses robust compression levels in ELA to minimise complexity and maximise accuracy.

DOI: http://doi.org/

 

 

Published by: vikaspatanker

Authors: Atharva Jadhav

Abstract: The advancement of artificial intelligence (AI) has given rise to two major approaches: traditional machine learning (ML) and deep learning (DL). While traditional ML relies on feature engineering and structured learning approaches, deep learning automates feature extraction through artificial neural networks. This paper explores the differences between these methods, compares their performance across domains such as image recognition, natural language processing, and financial forecasting, and evaluates their advantages and limitations. Experimental results and literature reviews indicate that deep learning excels in handling large datasets and complex patterns, whereas traditional ML is more suitable for smaller datasets with structured features.

 

 

Published by: vikaspatanker