Category Archives: Uncategorized

Authors: Kodavati Ram Sanjay, Ruban Kumar S, Saran Raj S, Dr. Arun Kumar

Abstract: Automated detection of crop leaf diseases and classification of remote sensing land cover categories remain challenging owing to complex backgrounds, illumination vari-ability, spectral distortions, and high intra-class visual similarity. Existing frameworks provide strong baselines but commonly suf-fer from scale-sensitive segmentation, redundant feature fusion, limited contextual representation, and slow convergence in high-dimensional feature spaces. This paper proposes the Adaptive Recurrent FusionNet (ARFusionNet) framework — a Flask-based web application that integrates four coordinated inno-vations: Multi-Scale Adaptive Contrast Normalisation (MACN) for illumination-robust preprocessing; Graph-Based Superpixel Attention Segmentation (GSAS) for adaptive Region-of-Interest extraction; Bidirectional Gated Recurrent Units (BiGRU) embed-ded within Residual Efficient Convolution Blocks with Adaptive Weighted Feature Aggregation (AWFA); and Hybrid Binary Differential Evolution controlled Particle Swarm Optimisation (BDE-PSO) for efficient feature selection. DenseNet121 serves as the backbone feature extractor. We validate the system on the Plant Pathology 2020 dataset (1,821 high-resolution apple leaf im-ages; four disease classes). ARFusionNet achieves 98.2% classifi-cation accuracy, surpassing the state-of-the-art baseline (97.6%), while reducing training time by approximately 78 seconds and remaining fully executable on a standard CPU laptop without GPU dependency. The accompanying web application exposes eight interactive diagnostic modules including leaf visualisation, Canny edge display, convolved feature maps, neural network architecture visualisation, and real-time per-image prediction.

Published by: vikaspatanker

Authors: Dr.T.V.Deokar, Namrata Manoj Bandgar, Ankita Dilip Thombare, Bhagyashri Ashok Bhong

Abstract: A Realtimeloadflowmonitorindistributionsystemthisprojectdevelopsareal-timeloadflowmonitoring system for a distribution network using a bulb and a motor. The system continuously monitors electrical parameters to detect overload and fault conditions. A GSM module is used to send instant alerts to the user, while a buzzer provides local warning. This ensures quick response, improved safety, and reliable operation. The project also demonstrates the effect of different load types onsystem performance and serves as a simple, cost-effective model for smart power distribution

Published by: vikaspatanker

Authors: Ashwani Kumar, Dr. Sunil Maggu

Abstract: The intersection of machine learning and preventive healthcare offers transformative potential for earlydisease detection and personalized health guidance. However, most existing systems either producebinary classification outcomes without contextual risk stratification or provide static, non-adaptivehealthrecommendations disconnected from individual prediction confidence scores. This paper introducesHealthGuard AI, a novel multi-stage predictive framework that integrates supervised machinelearningclassification with probability-based risk stratification and a dynamically adaptivehealthrecommendation engine. The system simultaneously addresses three major chronic diseasedomains—Type 2 Diabetes Mellitus, Coronary Heart Disease, and Parkinson’s Disease — using clinicallyvalidatedfeature sets drawn from UCI Machine Learning Repository datasets. Beyond binary prediction, HealthGuard AI applies predict_proba() outputs to stratify individual disease risk into Low(≤0.40), Medium (0.41–0.70), and High (> 0.70) categories, each triggering a distinct, evidence-alignedhealthrecommendation profile. An additional Body Mass Index and Basal Metabolic Rate estimationmoduleemploying the Mifflin-St Jeor equation further extends the system’s scope into nutritional healthanalytics. Deployed as an interactive web application via Streamlit Community Cloud, HealthGuardAIachieves classification accuracies of 78.5%, 81.3%, and 87.2% for Diabetes, Heart Disease, andParkinson’s Disease respectively. The system demonstrates that probability-aware risk stratification, when combined with adaptive, risk-tiered recommendations, produces a meaningfully richer andmoreclinically actionable output than conventional binary prediction pipelines. Experimental results, systemarchitecture, and the clinical relevance of risk-tiered adaptive recommendations are discussedindetail.

DOI: https://doi.org/10.5281/zenodo.19993977

Published by: vikaspatanker

Authors: Mahesh Ganji

Abstract: This research study examines incorporation of an AI based anomaly detection method of Oracle cloud Payroll to make the payroll more accurate, reduce risks, and enhance compliance. The performance of different machine learning models such as Isolation Forest, One-Class SVM, Neural Networks, and Logistic Regression is tested in terms of their performance in detecting payroll data anomalies. Findings indicate that models such as Logistic Regression performed moderately well but other models did not cope with false positives and poor anomaly detection. The future is to perfect the models, involve deep learning, and realize the real-time anomaly to make the payroll management in large organizations more efficient and accurate.

DOI: https://doi.org/10.5281/zenodo.19988211

Published by: vikaspatanker

Authors: Associate Professor Dr. S. Thaiyalnayaki, Janga Kishore, Kareti Manoj, Jogu Ganesh, Kasaragadda Gopi Chand

Abstract: The rapid growth of urbanization in metropolitan cities has significantly influenced real estate markets and housing prices. Accurately estimating property values has become increasingly important for buyers, sellers, and real estate investors. This study presents a machine learning-based house price prediction system designed to analyze housing data and estimate property prices based on multiple influential factors. The dataset used in this research includes property attributes such as location, square footage, number of bedrooms, and number of bathrooms collected from metropolitan regions including Chennai and Bengaluru. The proposed system applies data preprocessing techniques to improve the quality of the dataset before model training. These preprocessing steps include handling missing values, encoding categorical variables, and performing feature scaling to ensure consistent data representation. After preprocessing, a predictive model based on Linear Regression is implemented to analyze the relationship

DOI: https://zenodo.org/records/19981757

Published by: vikaspatanker

Authors: Prudvi Saisaran Ponduru

Abstract: Scalable artificial intelligence (AI) workflows increasingly fail not because individual models are weak, but because the surrounding architecture cannot process heterogeneous, bursty, high-stakes evidence at operational speed. This paper proposes ECHO-DR, an Event-Centric Hierarchical Orchestration architecture for real-time disaster response. The real-world problem addressed is the difficulty of turning social media, remote sensing, UAV imagery, weather alerts, seismic feeds, and incident reports into timely, auditable, and trustworthy operational intelligence during floods, earthquakes, wildfires, and storms. ECHO-DR introduces four core contributions: an event-centric memory plane that unifies vector retrieval, geospatial indexing, lakehouse lineage, and structured event graphs; a hierarchical routing policy that escalates only high-value or uncertain items to expensive multimodal reasoning; a stage-disaggregated serving design that independently scales encoders, prefill workers, decoders, and tool calls; and a governance plane that embeds auditability, human review, and zero-trust access control into the workflow. A formal utility-constrained routing model, event-linking algorithm, fusion rule, and capacity model are developed to show how the architecture scales under large workflows. The paper also provides an implementation blueprint, clean system diagrams, benchmarking methodology, ablations, and simulated evaluation results. Simulated trace-driven experiments indicate that the proposed gated architecture can reduce p95 provisional alert latency relative to a monolithic multimodal pipeline while maintaining evidence traceability and limiting deep-model cost. The work demonstrates that scalable AI for future big workflows should be designed as a compound, event-centered, policy-aware system rather than as a single model endpoint.

Published by: vikaspatanker

Authors: Asst. Prof. Rutuja Gautam, Prof. Rohan B. Kokate, St. Ankit R. Dhole

Abstract: Heart disease is one of the leading causes of death worldwide, making early prediction and diagnosis extremely important. This review paper focuses on the use of machine learning techniques for predicting heart disease based on medical data. Various algorithms such as Logistic Regression, Decision Tree, Random Forest, and Support Vector Machine are analyzed for their effectiveness in prediction. The system uses patient health parameters like age, blood pressure, cholesterol level, and heart rate to determine the risk of heart disease. A web-based application is also discussed, developed using Python for backend processing and HTML/CSS for user interaction. The results show that machine learning models can significantly improve prediction accuracy and assist doctors in decision-making. This paper highlights the importance of data preprocessing, model selection, and performance evaluation in building an efficient heart disease prediction system.

DOI: https://doi.org/10.5281/zenodo.19975051

Published by: vikaspatanker

Authors: Vishal Tandale, Shabbir Ahmed, Hitesh Shewale

Abstract: The dairy sector remains a cornerstone of Indian agriculture, facing persistent challenges such as manual inefficiency, data fragmentation, and increasing demands for quality and traceability. This paper analyzes contemporary dairy management systems, focusing on the adoption of digital technologies—Internet of Things (IoT), Artificial Intelligence (AI), and integrated Enterprise Resource Planning (ERP)—to streamline operations, optimize productivity, and improve animal welfare. Evidence from recent deployments and technology pilots demonstrates that technologically- augmented management not only boosts efficiency but also aligns the sector with Food Safety and Standards Authority of India (FSSAI) compliance and export requirements. Implementation challenges and recommendations for scalable, farmer-friendly solutions are discussed.

Published by: vikaspatanker

Authors: Vrushabh Jitendra Patil, Pratik Prakash Patil, Ganesh Rajendra Mote

Abstract: Traditional irrigation systems depend heavily on manual inspection to detect leaks, pipe blockages, and abnormal water flow, resulting in significant water wastage, reduced irrigation efficiency, and increased maintenance expenditure. This paper presents AquaVision BI, an intelligent IoT-enabled irrigation monitoring system that integrates three Hall-effect flow-rate sensors, an ESP32 Wi-Fi microcontroller, and an AI-driven differential-threshold anomaly-detection algorithm to achieve real-time surveillance of irrigation pipelines. The system continuously samples sensor pulse counts at one-second intervals, computes volumetric flow rates, and applies pairwise differential analysis to localise leakage to specific pipeline segments (upstream, mid-stream, or downstream). Upon anomaly detection, automated alerts are dispatched via the Blynk IoT cloud dashboard and a local buzzer actuator. Experimental evaluation on a controlled testbed confirms accurate leak localisation across all three sensor nodes, with end-to-end alert latency consistently below two seconds. The proposed system significantly reduces water wastage, lowers operational costs, and promotes sustainable agricultural water management practices.

Published by: vikaspatanker

Authors: Khushi, Rajat Takkar, Mugdha, Himanshi, Muskan

Abstract: Schools are embracing the use of data-driven information to track student achievement and performance. Conventional ways of tracking performances are not effective in the intricate nature of the relationships among diverse academic contributions. This research works on the necessity of an efficient, but simple, artificial intelligence based framework to examine and visualize the factors mentioned and to take proactive measures that will help find out students who might not need more than a top-quality academic assistance. The main purpose of the research is to come up with a machine learning model that is easy to interpret, has the predictive strength of the end-of-year academic scores and classifies students into groups of "pass" and "fail." Also, the research will visualise the relationship between particular inputs (ex: hours of study, attendance) and performance in general and characterize a feature importance analysis to determine which factors have the most profound impact on student achievement. The research works with the artificial data including major academic variables: study hours, attendance, assignment grades, internal grades, and past GPA. The methodology will use two different machine learning models: Linear Regression to predict continuous performance scores and Decision Tree Classification to perform binary categorisation (Pass/Fail). Visualisation tools were combined to plot interactions among variables, and parameter analysis was performed in terms of standard accuracy measurements of regression as well as classification problems. The results indicate that the most important predictors of academic success are study hours, internal marks and attendance. Linear Regression model largely was able to predict final scores with high correlation to input data whereas the Decision Tree classifier offered a simple, interpretable logic with which students can be categorised. Analysis of feature importance provided a reason on why the consistent engagement and incremental assessment has a greater influence on the outcome rather than just the previous GPA. The offered AI-based framework is a scalable and understandable research proposal method of analysing educational data. The system facilitates informed, data-driven decisions made by educators by highlighting its critical performance drivers, and it helps to deliver timely interventions to at-risk students. Further work would entail the application of the framework on bigger datasets, and real-life contexts to improve predictive accuracy in diverse education settings.

DOI: https://doi.org/10.5281/zenodo.19960357

Published by: vikaspatanker