Daily Archives: July 2, 2025

Authors: Mrs. Penki Tulasi Bai, Mrs. P. Manasa

Abstract: This study suggests employing predictive maintenance to enhance the operational efficiency and prolong the lifespan of industrial machinery and equipment through machine-learning techniques. As producers prioritize reducing downtime and cutting expenses, proactive maintenance strategies are becoming increasingly vital for ensuring operational reliability. The research aims to gather historical data to train machine-learning models that can predict equipment failures and develop an algorithmic framework for scheduling preventive maintenance. The primary objective is to assist in forming an effective anticipatory maintenance strategy, which can lower industrial maintenance costs and improve product prices. Various machine-learning techniques, along with extensive data preprocessing and feature engineering methods, will be utilized in this research. Data preprocessing will involve tasks such as cleaning, dataset conversion, and normalization prior to model training. Feature engineering will focus on identifying the most important characteristics for accurate prediction of machine failures. Numerous machine-learning methods, including Random Forest (RF), Long Short-Term Memory (LSTM), and Support Vector Machines (SVM), will be evaluated to determine the most effective model for precise forecasting. The performance of these models will be compared using metrics such as Root Mean Square Error (RMSE), R-squared (R²), and Mean Absolute Error (MAE) as indicators. Ultimately, the top-performing machine-learning models will be integrated into real industrial settings, with the optimal model expected to achieve a 5-10% increase in operational efficiency.

DOI: http://doi.org/10.5281/zenodo.15790457

 

Published by: vikaspatanker

Authors: Ms. Bobbili Bhargavi, Mrs. K. Sowjanya

Abstract: The complex evolution of railway cars influences transit routes. Many mistakes arise from the utilization of train lines. Both Manufacturing mistakes and improper rail usage are responsible. For these deficiencies. There are numerous techniques for detection. Errors must be recognized promptly and rectified. The camera-based technique is One of these methods. By utilizing cameras affixed to the railway vehicle, images of the rail components are examined. Flaws are identified in the rail components. A method for detecting and analysing defects in rail tracks. Surfaces are proposed in this document. The recommended method employs image processing to identify the rail surface. High resolution images captured by specialized cameras mounted on the proposed system encompasses railway inspection cars. A Variety of track issues, including cracks, weld defects, and track Misalignment and ballast degradation are detected. These images were utilized to perform an analysis. Pre-processing and feature extraction. Image processing entails the application of segmentation techniques. Procedures to isolate the track area and emphasize any Potential defects. Fuzzy logic is employed to prioritize maintenance tasks. Based on urgency once issues have been identified and their Severity has been evaluated. Fuzzy logic is particularly adept at capturing the subjective assessments involved in evaluating. track conditions as it offers a flexible framework. Processing ambiguous and imprecise data. To assign appropriate severity ratings for the identified features of each issue. Type, fuzzy rules, and membership functions are developed.

DOI: http://doi.org/10.5281/zenodo.15790262

 

Published by: vikaspatanker

Authors: Ms . Nellipudi Sai Sravani1, Dr Sivabalan Settu Ph.D, Postdoc 2

Abstract: Despite the numerous advantages that electric vehicles (EVs) offer in terms of environmental protection and emission reduction, their widespread acceptance is primarily influenced by their pricing. By utilizing machine learning (ML) algorithms, it is possible to forecast these costs. This study seeks to evaluate the effectiveness of several prominent ML algorithms to ascertain which one is most capable of accurately predicting the prices of electric vehicles. In order to pinpoint the essential factors, we conducted a literature review to investigate the elements that influence the pricing of electric vehicles, facilitating our cost estimation. We theoretically assessed these ML algorithms to corroborate our results and subsequently compared the findings of this comparative analysis with the results obtained from the simulations.

DOI: https://doi.org/10.5281/zenodo.15790332

 

Published by: vikaspatanker

Authors: Mrs . Kolli Kundana Bhavya Sree, Mrs. B Sirisha, Mtech, Associate Professor

 

 

Abstract: To ensure the reliability of connected devices, machine learning is employed to analyse network traffic, facilitating quicker identification of unusual behaviour and congestion. The application of machine learning methods improves the ability to manage traffic and supports the maintenance of service quality. Furthermore, the role of machine learning in network security is to identify anomalies and classify traffic in real-time, aiming to optimize network performance and uncover potential threats. This study highlights the beneficial effects of utilizing machine learning techniques to improve network reliability and security. One of our contributions is an examination of an example of HTTP/3 traffic interacting with a web server. We implemented machine learning algorithms to differentiate between standard traffic and possible HTTP/3 flood attacks. Additionally, we developed a dataset of traffic samples featuring 23 attributes categorized into six subgroups. From traffic captured in a simulated environment, we evaluated the significance of these attributes and discovered that employing machine learning techniques can greatly enhance both network security and reliability. We utilized four supervised classification algorithms: Support Vector Machine (SVM), Logistic Regression, Random Forest, and K-Nearest Neighbours (KNN). These algorithms represent a category of supervised classification methods. They played a crucial role in training datasets of network traffic, which were carefully labelled to distinguish between Distributed Denial-of-Service (DDoS) attacks and normal traffic. The results of this research demonstrate the efficacy of machine learning algorithms in analysing network traffic to detect specific types of DDoS attacks, especially those that use QUIC traffic. This illustrates the significant potential of machine learning techniques in strengthening the overall security and reliability of networks.

DOI: http://doi.org/

 

 

Published by: vikaspatanker

Authors: Ms. Manepalli Kavya, Mrs. Jitendar Ahuja

Abstract: Recent years have seen a significant increase in phishing attacks targeting websites, posing persistent challenges to digital security. While numerous detection tools have been developed, they often fall short in comprehensively identifying all threats and struggle with subtle, evolving forms of deception. Integrating machine learning (ML) techniques has emerged as the most effective strategy to overcome these limitations, significantly enhancing detection accuracy and computational efficiency. This approach is crucial for addressing the shortcomings of existing phishing detection models. This paper introduces an Intelligent Phishing defence paradigm, leveraging an ENASSEMBLE-driven ML model specifically trained on a designed dataset for high-fidelity website identification. Our objective is to demonstrate how the ENASSEMBLE model not only bolsters the overall accuracy of phishing detection but also offers a robust and efficient solution capable of recognizing complex and evasive fraudulent sites, thereby fortifying online security.

DOI: https://doi.org/10.5281/zenodo.15790035

 

Published by: vikaspatanker

Authors: Mr. Battu Rajesh, Associate Professor Mr. M. Satyanarayana

Abstract: In this study, a content-based image retrieval (CBIR) system was built as an efficient image retrieval tool, allowing the user to send a query to the system, which then retrieves the user's desired image from the image database. We wanted to present a quick overview of the novel coronavirus (SARS-CoV-2) and a better knowledge of the coronavirus illness (COVID-19) in diabetics and its therapy. In this study, we use the COVID-19 dataset to train machine learning algorithms, which subsequently predict whether a person has type diabetes. If type 2 diabetes is detected in a person's test record, he is more prone to COVID-19 disease, heart disease, or renal disease.

DOI: https://doi.org/10.5281/zenodo.15789601

Published by: vikaspatanker

Authors: Mr.Gurudas jadhav, Mr. Mayur Shinde, Dr. Jasbir Kaur, Assistant Professor Mr. Suraj Kana

Abstract: With the explosion of data in recent years, the methods used for extracting, transforming, and loading (ETL) or extracting, loading, and transforming (ELT) data have become critical in the design of modern data pipelines. These methodologies are pivotal in ensuring that raw data from disparate sources is cleansed, structured, and made analytics-ready for business decision-making and operational insight. The efficiency and effectiveness of these processes directly impact the performance of data warehouses and the value extracted from data analytics initiatives. This paper presents a comprehensive comparison of ETL and ELT paradigms in terms of architecture, performance, scalability, cost efficiency, governance, and use-case suitability. Through an in-depth exploration of their underlying technologies, application scenarios, and industry adoption patterns, we aim to clarify the decision-making process for choosing the right approach in different organizational contexts. We consider technical, operational, and business dimensions that influence the selection between ETL and ELT, including data volume, regulatory compliance, tool ecosystems, and team skillsets. Moreover, we delve into the role of emerging cloud-native platforms that support ELT’s rise, and how modern engineering practices such as version control, CI/CD, and modular design are redefining data transformation workflows. Case studies from leading technology firms illustrate practical implementations and benefits of these approaches, highlighting real-world trade-offs. We also explore the future trends and hybrid architectures that aim to harness the strengths of both paradigms in increasingly complex data environments, particularly in light of advancements in artificial intelligence, real-time processing, and decentralized data ownership models such as data mesh. By synthesizing insights from academic research, industry white papers, and technical documentation, this paper provides a strategic framework to guide enterprises in architecting resilient, scalable, and future-ready data integration solutions. The paper concludes with references to academic research, industry white papers, and technical documentation.

 

 

Published by: vikaspatanker

Authors: Rutik Jadyar, Hritik Acharya, Dr. Jasbir Kaur, Assistant Professor Ms. Ifra kampoo

Abstract: In recent years, the use of digital health data has grown rapidly. However, sharing sensitive medical information can lead to serious privacy concerns. Traditional data analysis methods require centralizing data, which poses a risk of exposing private information. Federated Learning (FL) is a new method that allows hospitals and healthcare institutions to collaborate on machine learning models without sharing actual patient data. Instead, the model is trained across different devices or servers holding local data. This paper explains how FL works, its benefits for healthcare, and how it can be applied to protect patient privacy while still enabling powerful data analysis.

 

 

Published by: vikaspatanker