Category Archives: Uncategorized

Authors: Deepa Barethiya, Dhanashri Pannase, Gangasagar Kashyap

Abstract: Ensemble machine learning has pushed crop recommendation accuracy past 99% on standard soil-weather benchmarks — yet this milestone conceals a troubling gap. Systems built around Random Forest, XGBoost, and gradient boosting produce ranked crop labels while leaving the economic viability of each suggestion entirely unexamined. A farmer told "grow rice with 99% confidence" still does not know whether that choice will leave a positive margin after seed, fertiliser, and irrigation costs. This paper proposes a decision-oriented framework that moves beyond the accuracy plateau by coupling a soft-voting ensemble with per-crop yield regressors and a configurable economic layer that estimates expected profit. Where conventional pipelines terminate at a suitability label, the proposed architecture extends the output to a Risk-Adjusted Expected Profit, mathematically formulated as E[Π_c ]_(risk-adjusted)=P_ensemble (c│X) Π_c, where P_ensemble (c│X)is the Ensemble Suitability Probability and Π_c=((Y_c ) ̂(X)×P_(market,c)×1000)-Total Cost_cis the Nominal Net Profit. This coupling mathematically discounts the apparent value of high-risk crops by their probability of soil-weather failure — a correction absent from every reviewed system. To illustrate the theoretical decision dynamics of this framework, we construct a conceptual walkthrough across 200 hypothetical soil-weather scenarios derived from standard agricultural benchmarks. This analysis suggests that the agronomically top-ranked crop and the economically top-ranked crop diverge in roughly 46% of cases — a finding that, if borne out in empirical deployment, would have direct implications for farm-level income planning. A conceptual Streamlit dashboard design is also proposed, embedding real-time what-if sliders and SHAP-based feature attributions to make the system transparent to extension workers and farming cooperatives. The central argument of this paper is simple: a classifier that ignores profit is only half a tool. This framework proposes the other half.

DOI: https://doi.org/10.5281/zenodo.20927114

Published by: vikaspatanker

Authors: Assistant Professor Deepa Barethiya, Sakshi Jibhkate, Samiksha Daronde

Abstract: This paper compares Java-based frameworks and ASP.NET Core for web applications used by companies. It looks at how they work and how well they handle a large number of users. The study checks things like how long it takes for the application to respond, how much work it can handle, how much of the computer’s brain it uses and how much memory it needs when a lot of people are using it at the same time. They ran tests to see what would happen if many people used the application. The results show that ASP.NET Core is really good at responding and using resources wisely. Java-based frameworks are good at handling a lot of users and working with computers at the same time. This study tells us what is good and what is not so good about Java-based frameworks and ASP.NET Core. It helps people choose the tools to build big web applications for companies.

DOI: https://doi.org/10.5281/zenodo.20926193

Published by: vikaspatanker

Authors: Deepa Barethiya, Kshitij Moon, Dipak Meshram

Abstract: Accurate Remaining Useful Life (RUL) prediction for turbofan engines is critical for implementing effective condition-based maintenance strategies, enhancing operational safety, and reducing maintenance costs. Traditional predictive maintenance approaches often struggle with the non-linear, time-dependent characteristics of engine degradation. This paper presents a data-driven prognostic model utilizing a Long Short-Term Memory (LSTM) neural network to predict the RUL of turbofan engines based on sensor-derived operational data. The model is trained and validated on the NASA C-MAPSS dataset, which contains run-to-failure data for multiple turbofan engines. The proposed methodology involves preprocessing raw sensor data, creating sequential inputs using a sliding window approach, and training a two-layer LSTM architecture designed to learn complex temporal degradation patterns. Model performance is evaluated using standard regression metrics, including Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and R² score. The resulting model demonstrates robust predictive capabilities and is deployed in a Flask-based web application, offering a practical tool for real-world CBM systems and highlighting the efficacy of deep learning for industrial prognostics.

DOI: https://doi.org/10.5281/zenodo.20925860

Published by: vikaspatanker

Authors: Deepa Barethiya, Himanshu Praveen Dethekar, Bhavesh Tembhurkar

Abstract: The dark web constitutes a stratified, operationally sophisticated cybercrime ecosystem whose threat dynamics are shaped by layered anonymity infrastructure, AI-augmented criminal tooling, and resilient financial obfuscation mechanisms. While existing literature provides valuable but fragmented analysis of individual components, few studies integrate these elements within a unified analytical framework. This paper addresses that gap through a hybrid analytical survey approach, advancing four primary contributions: (1) a six-dimension taxonomic model differentiating surface web, deep web, and dark web environments; (2) a Five-Layer Dark Web Threat Ecosystem Model characterising the functional architecture of criminal infrastructure; (3) a structured capability taxonomy of AI-augmented criminal tools (Dark LLMs); and (4) a proposed Cyber Threat Intelligence (CTI) extraction pipeline for dark web environments. Drawing on peer-reviewed literature spanning 2020–2025, operational intelligence from Europol IOCTA, Chainalysis Crypto Crime Reports, and FBI IC3 data, and documented threat actor behaviour, the paper analyses ransomware-as-a-service dynamics, cryptocurrency financial obfuscation, law enforcement response limitations, and post-Tor architectural evolution. Persistent research gaps in multilingual CTI extraction, post-Tor forensic methodology, and AI-threat detection are identified, with a structured research agenda proposed.

DOI: https://doi.org/10.5281/zenodo.20925387

Published by: vikaspatanker

Authors: Anuradha Muttamwar, Esha Dorkhande, Vaibhavi Meshram

Abstract: The exponential proliferation of digital media in the modern era has created an environment where mis- and disinformation as well as "fake news" can spread uncontrollably, leading to challenges to public discourse, political trust and integrity. In this paper we present a detailed research approach toward fake news detection through efficient feature engineering and the use of supervised machine learning. We use a dataset composed of 5,000 current news articles (2,537 real, 2,463 fake news) and conduct an in-depth research regarding the performance of TF-IDF with n-grams. We build and train a Multinomial Naive Bayes model and attain excellent classification accuracy. Furthermore, we investigate the importance of text preprocessing such as stop word removal, stemming and lemmatization. Our model achieves a final accuracy of 93.6%, while also achieving scores for precision, recall and F1 greater than 0.92. When comparing with baseline models, the presented method with enhanced feature engineering shows excellent results. We then developed a web based system with the help of Flask that allows real time fake news detection and confidence. It will establish a reusable, light and scalable pipeline to automate fake news detection in real world applications.

DOI: https://doi.org/10.5281/zenodo.20924573

Published by: vikaspatanker

Authors: Anuradha Muttamwar, Bhumika Balpande, Shantanu Gawai

Abstract: Skin diseases are a major health concern worldwide, but getting an appointment with a dermatologist can be tough, especially in rural areas. That's why we've created a web-based system that uses artificial intelligence to help diagnose skin conditions. Our system is built using the Node.js framework and combines a powerful image classification model with a user-friendly website. Here's how it works: users upload pictures of their skin through a simple interface, and our system uses a special kind of neural network called a Convolutional Neural Network (CNN) to analyze the image and make a prediction. We've trained our model using a technique called transfer learning, which allows it to learn from existing knowledge and apply it to new situations. Our model can accurately diagnose five common skin conditions: eczema, acne, psoriasis, dermatophytosis, and benign nevi. We've designed our system to be fast and efficient, even when lots of people are using it at the same time. Our tests show that it can handle up to 100 users simultaneously without slowing down, and it can give results in under a second. We're excited about the potential of our system to provide a low-cost, accessible way for people to get a preliminary diagnosis and take the first step towards getting treatment. Our system is made up of three main parts: a website that users interact with, a backend server that handles the image analysis, and a database that stores all the information.

DOI: https://doi.org/10.5281/zenodo.20924037

Published by: vikaspatanker

Authors: Mayuri Dongre, Harsh Upase, Krushnakant Shinde

Abstract: Artificial Intelligence (AI) has emerged as a transformative technology in modern agriculture, enabling sustainable and data-driven farming practices through precision agriculture techniques. This research paper presents a comprehensive study of AI-based technologies and their applications in sustainable precision agriculture. The study explores the integration of Machine Learning (ML), Deep Learning (DL), Internet of Things (IoT), computer vision, robotics, drones, and sensor-based systems for improving agricultural productivity, resource optimization, and environmental sustainability. AI techniques are increasingly used for crop prediction, disease detection, soil analysis, irrigation management, yield forecasting, weed identification, and climate monitoring, helping farmers make accurate and timely decisions. The paper also highlights how precision agriculture minimizes the excessive use of water, fertilizers, and pesticides while enhancing crop quality and reducing environmental impact. Furthermore, the study examines recent advancements, real-world applications, challenges, and limitations. AI adoption in agriculture, including high implementation costs, lack of technical knowledge, data availability issues, and infrastructure constraints in rural areas. Precision agriculture harnesses data-driven techniques to optimize crop production, resource use, and sustainability. However, low-income countries like Bangladesh face a short- age of localized, high-quality datasets that reflect regional agroclimatic conditions and cropping practices.

DOI: https://doi.org/10.5281/zenodo.20923886

Published by: vikaspatanker

Authors: Anuradha Muttamwar, Devashri Ghotekar, Damini Mishra

Abstract: In the contemporary digital landscape, social media has become an integral part of daily communication for billions of users worldwide. While these platforms facilitate connectivity and self-expression, the growing trend of oversharing personal information has created unprecedented cybersecurity and privacy risks. Users increasingly disclose sensitive information such as location details, financial data, personal relationships, and health conditions, often without fully comprehending the potential consequences. This research presents a comprehensive study on oversharing culture, examining how habitual social media usage patterns intensify individual vulnerability to identity theft, social engineering attacks, data breaches, and psychological manipulation. The study integrates behavioral analysis, cybersecurity assessment frameworks, and vulnerability evaluation metrics to understand the mechanisms driving oversharing behavior and its security implications. Through survey-based analysis and comparative study of social media platforms, we examine the psychological motivations behind excessive self-disclosure, including the role of social validation through likes and comments, platform design strategies, and individual personality traits. The research demonstrates that approximately 93% of users who overshare personal information face significant privacy and security risks, making vulnerability assessment and user education critical priorities. The proposed framework employs data analysis techniques, behavioral pattern recognition, and machine learning algorithms to identify vulnerability indicators and predict susceptibility to cyber threats. The visualization layer presents findings through interactive dashboards and heat maps, enabling users and security professionals to understand oversharing risks and implement protective measures. Our findings indicate that comprehensive awareness programs, behavioral intervention strategies, and platform-level privacy controls can significantly reduce vulnerability when combined with individual digital literacy initiatives.

DOI: https://doi.org/10.5281/zenodo.20923714

Published by: vikaspatanker

Authors: Gawai Santosh Bhaskar, Shashikant B. Dhobale

Abstract: The process of developing a comprehensive project cost estimate is critical for a project to be adjudged successful on completion. Projects’ costing is one of the most critical and most widely used project management tools. The complex nature of Projects and the inherent uncertainty of the financial performance of construction projects, development funding, and the monitoring and controlling of costs and schedules make exact budget needs impossible to forecast accurately. This same characteristic also makes projects to deviate from plans. The main object of this paper is to identify the factors affecting the accuracy of project cost estimation, determine the various methods of carrying out project cost estimation in construction projects within INDIA. The study is motivated by the inability of most construction professionals to arrive at a tentative and reliable project cost estimate in project realization which has created obvious problems of project cost overrun and subsequent abandonment. The study sampled the opinion of fifty-three selected project professionals who had worked on related construction outfits in INDIA. An objective realization instrument developed using eighteen (18) factors identified in the literature as possible factors affecting the accuracy of project cost estimation were ranked based on a Likert four-point scale. The score of respondents to the factors were analyzed using descriptive and inferential statistics, mean score value and factor analytical approach as the major tool.

Published by: vikaspatanker

Authors: Rakesh Dondapati

Abstract: Corporate environmental, social, and governance (ESG) disclosures increasingly rely on persuasive sustainability narratives, yet investors, regulators, and civil society organizations often lack scalable tools to distinguish genuine environmental performance from rhetorical positioning. This study develops and validates a Greenwashing Intelligence System (GIS) that integrates six data modalities — ESG narrative text, verified emissions data, satellite and remote-sensing indicators, controversy and incident records, financial disclosures, and supply-chain risk signals — to construct two independent indices: a Narrative Ambition Score (NAS), derived from transformer-based analysis of sustainability disclosure text, and a Performance Index (PI), derived from verified and independently observable environmental performance data. The difference between these indices, the Greenwashing Gap Score (GGS = NAS – PI), is computed for a global panel of 4,642 public firms across five regions and six sectors over a 2019–2026 observation period. Firms are classified into four quadrants: Aligned Leaders (high NAS, high PI, 23.5% of sample), Greenwashing Risk (high NAS, low PI, 16.0%), Quiet Achievers (low NAS, high PI, 13.3%), and Disengaged (low NAS, low PI, 31.7%). Regression results show that GGS significantly predicts negative cumulative abnormal returns around disclosure events (β = –0.041, p < .001), elevated 24-month litigation risk (β = 0.0021, p < .001), and negative media sentiment shifts (β = –0.0089, p < .001), with these relationships substantially amplified when satellite-reported divergence (SRD) is high (GGS × SRD interaction significant across all outcomes, p < .001) — indicating that externally verifiable narrative-performance gaps carry the largest market and reputational consequences. Sector analysis reveals the largest gaps in Energy and Materials sectors, particularly for Scope 3 emissions claims. A validation study comparing GIS classifications against a 180-member expert panel shows substantial agreement (Cohen's κ = 0.65–0.78 across classification dimensions). A two-year disclosure-change pilot demonstrates that sharing GIS reports with firms reduces subsequent GGS, with the largest reductions (–9.7 points) among Greenwashing Risk firms receiving publicly benchmarked reports. The paper contributes the GIS architecture, the NAS/PI/GGS measurement framework, and a five-level ESG assurance maturity roadmap to ESG analytics, accounting information systems, and AI governance research, demonstrating that multimodal AI can operationalize sustainability assurance at scale.

DOI: http://doi.org/10.5281/zenodo.20920915

Published by: vikaspatanker