Authors: Indhuprakash M, Gokul R, Harishragavendra M, Parthasharathy R, M. Premkumar

Abstract: Weather prediction supports agriculture, transportation, travel planning, and public safety. This paper presents an Intelligent Weather Prediction System that combines machine learning techniques with an accessible application interface. The proposed workflow uses weather observations such as temperature, humidity, atmospheric pressure, wind speed, cloud cover, and rainfall, depending on the data source. It includes data collection, preprocessing, exploratory analysis, feature preparation, model training, evaluation, and presentation of predictions. Candidate models include regression methods for continuous weather variables, Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Long Short-Term Memory (LSTM) for suitable classification, regression, or sequential forecasting tasks. The application may also provide AI-assisted weather question answering and concise weather-news summaries when these components are implemented and connected to reliable sources. The system is intended to make weather information easier to understand and use. Actual performance must be measured on the selected dataset; no accuracy result is claimed in this draft. The final paper should report the implemented models, dataset, evaluation results, and system screenshots

DOI: https://doi.org/10.5281/zenodo.23586599

Published by: vikaspatanker

Authors: Dinesh G, Kabilan KM, Sowndar K, Vikram S, Indhuja N

Abstract: Student performance prediction is an important application of machine learning in educational data mining. Academic institutions collect different types of information, including attendance, internal assessment marks, previous academic performance, study habits, and assignment completion. Analyzing these factors manually can be time-consuming, particularly when the number of students increases. This paper presents a machine-learning-based framework for predicting student academic performance using relevant educational attributes. The proposed system consists of data collection, data preprocessing, exploratory data analysis, feature selection, model training, evaluation, and prediction. Classification algorithms such as Logistic Regression, Decision Tree, Random Forest, and Support Vector Machine can be evaluated to identify a suitable model for categorizing students into performance levels. Regression methods can alternatively be used when the target variable is a numerical examination score. The system is intended to help educators identify students who may require additional academic support and plan appropriate interventions. Model performance is evaluated using appropriate classification or regression metrics. Since actual experimental measurements depend on the selected dataset and implementation, the proposed framework does not assume unverified accuracy values. The study also discusses data privacy, fairness, interpretability, and the limitations of predictive educational systems.

DOI: https://doi.org/10.5281/zenodo.23590176

Published by: vikaspatanker

Authors: Wani Ishwari Dnyaneshwar, Professor Abhishek Popat Bangar, Professor Bhosale S. B., Dr. A. A. Khatri

Abstract: Cross-border trade remains costly and uneven. Exporters, especially small firms in developing regions, face language barriers, limited market knowledge, complex customs and compliance procedures, fragmented logistics and unequal access to digital tools. Artificial intelligence (AI) is widely proposed as a way to lower these costs and to improve global marketing through translation, demand forecasting, segmentation, personalisation and process automation. This paper presents a focused narrative review that synthesises five sources: a causal field study of machine translation on an online trade platform, a strategic framework for AI in marketing, a bibliometric review of 1,580 AI-in-marketing papers, a systematic review of 64 studies of AI in supply chain management, and the World Trade Organization’s World Trade Report 2025. The review examines three connected themes: AI techniques and marketing frameworks relevant to trade, the strength of the empirical evidence, and the conditions that matter for deployment by small exporters. The only causal study reports that a machine translation upgrade raised exports on eBay by 10.9%, which its authors describe as equivalent to reducing the distance between trading countries by 26.1%. The WTO projects that AI could raise cross-border trade in goods and services by nearly 40% by 2040, but only with enabling policies, and warns that unequal access could widen divides. However, the technical literature is dominated by model-building, few studies test usability or generalisability, and most are not specific to an industry or country. The review contributes a comparative table, a conceptual architecture for an AI-enabled export platform, and a prioritised agenda of research gaps.

DOI: https://doi.org/10.5281/zenodo.23566631

Published by: vikaspatanker

Authors: Vaishnavi Vadane, Abhishek P. Bangar, Sachin B. Bhosale, V. A. Karad

Abstract: Industrial production depends on electricity, process heat, motors, pumps, compressed-air networks, refrigeration, and production schedules. Artificial Intelligence (AI) can help identify avoidable energy use and support decisions that reduce associated greenhouse-gas emissions. This review examines Green AI strategies for industrial energy and carbon reduction, including energy forecasting, predictive maintenance, process optimisation, intelligent scheduling, digital twins, renewable-energy coordination, and carbon-aware operation. It also considers the environmental cost of AI itself, such as energy used for model training and inference, data storage, networking, and hardware. The reviewed literature indicates that AI-enabled monitoring, prediction, control, scheduling, and parameter optimisation are important application areas. However, reported benefits depend on data quality, system integration, baseline selection, operator acceptance, and verification in real operating conditions. The central research gap is the lack of a consistently applied, independently verifiable framework that measures net energy and carbon benefits across the industrial system while accounting for the AI system’s own footprint and comparing AI with conventional engineering alternatives. This paper organises the gaps, proposes research questions, and presents a practical framework for future work.

DOI: https://doi.org/10.5281/zenodo.23566196

Published by: vikaspatanker

Authors: Swara Sandip Lende, Professor Bangar Abhishek Popat, Professor Bhosale S.B.

Abstract: Artificial intelligence (AI) is increasingly used to accelerate scientific discovery, from predicting protein structures and molecular properties to extracting knowledge from the scientific literature. A large share of experimental evidence in papers is reported in numerical tables, yet automated analysis of such tables remains difficult. This paper presents a focused narrative review of verified sources on AI for scientific discovery, concentrating on large language model (LLM) based understanding of scientific text, scientific tables and numerical reasoning. It examines three connected themes: knowledge extraction and question answering over scientific text, table question answering benchmarks, and reasoning methods (direct prompting, chain-of-thought, program-of-thought, self-consistency and graph-based reasoning). The literature shows that scientific table benchmarks have moved from mostly lookup tasks towards a 1,607-question benchmark that requires multi-step numerical reasoning; on it, GPT-4.1 mini reached only 27.0 exact match under direct prompting on the computation-and-language subset, chain-of-thought prompting raised this to 71.5, and accuracy still fell from 74.6% on small tables to 59.2% on large ones. However, the evidence is concentrated in computer science tables, depends on LLM-generated and LLM-filtered questions, and agreement among human verifiers was only moderate (Fleiss' kappa 0.55). The review contributes a comparative analysis of benchmarks and methods, a proposed conceptual architecture linking extraction, cell selection, stepwise reasoning, verification and expert review, and a prioritised list of research gaps.

DOI: https://doi.org/10.5281/zenodo.23565700

Published by: vikaspatanker

Authors: Sonawane Shubhangi Chintaman, Bangar Abhishek Popat, Bhosale Sachin Bajirao

Abstract: Healthcare organisations are exploring AI systems that do more than answer questions: they plan multi-step tasks, retrieve evidence, call software tools and coordinate with other agents. This capability promises support for clinical decision-making and administration, yet it widens the security and safety exposure of systems that handle highly sensitive patient data. This paper presents a structured narrative review of five verified IEEE journal articles published in 2025–2026. They cover the foundations of agentic AI, a taxonomy-based survey of healthcare agents, a multi-agent clinical decision support system, a multi-agent framework for Alzheimer's disease analysis and a privacy-preserving language-model framework for medical consultation. The papers are compared on objectives, methods, contributions, limitations and relevance to secure healthcare. The evidence indicates that retrieval-grounded, multi-agent designs are the dominant pattern, whereas human-in-the-loop control, adversarial testing, drift monitoring and regulatory compliance are rarely implemented. In the parts of the sources that could be read, no prospective clinical outcome trial is reported. The paper contributes a thematic synthesis, a five-study comparison, a conceptual layered architecture assembled from safeguards found in the literature, and a research agenda. The review is deliberately narrow and makes no claim of systematic coverage.

DOI: https://doi.org/10.5281/zenodo.23565555

Published by: vikaspatanker

Authors: Shrikant Patne, Professor Bangar Abhishek Popat, Professor Bhosale Sachin Bajirao

Abstract: Food that looks fine on the shelf is not always safe to eat, yet the decision is still made, for the most part, by human inspectors working under time pressure. This paper presents a machine-learning pipeline that grades fresh produce into three quality classes (fresh, medium and spoiled) from ordinary RGB images. We compare three classical learners (a support vector machine, k-nearest neighbours and a random forest) built on hand-crafted colour and texture descriptors with three deep models: a lightweight custom convolutional network, ResNet50 and a fine-tuned MobileNetV2. On a balanced dataset of 7,450 images, the fine-tuned MobileNetV2 reached 96.7% accuracy and a macro F1-score of 96.7%, which is 7.6 percentage points above the best classical baseline, while staying small enough (2.6 million parameters) to run on a mid-range smartphone CPU in roughly 38 ms per image. Ablation experiments indicate that data augmentation and transfer learning account for most of the improvement, and Grad-CAM maps confirm that the network focuses on decay-related regions such as dark patches and wrinkled skin. We also discuss where the approach still struggles, mainly the borderline medium class, and outline a route toward sensor fusion for real-world deployment.

DOI: https://doi.org/10.5281/zenodo.23565257

Published by: vikaspatanker

Authors: Shinde Bhumika Haridas, Professor Bhosale S.B, Professor Bangar A.P, Professor Khatri A.A

Abstract: Scientific research usually begins with a question, followed by a review of existing work, the development of possible The reasons for this, the planning of the experiments or simulations, and a careful verification of the results. Now artificial intelligence (AI) is being used can be used to help with a number of these steps. Large language models (LLMs) assist researchers in working with scientific information, AI agents can Tasks can be carried out using external tools, and evolutionary methods can continuously generate and test programs. The paper looks at four studies. that look at different parts of AI-assisted discovery: the wider influence of AI on science, autonomous agent systems, the TeLLAgent and the use of large language models in the search for evolutionary programming. Together, these studies demonstrate how AI could be of assistance with literature reviews and idea generation. They also highlight important issues, such as incorrect model outputs, and involve generation, computational testing, and program improvement. data that is biased or incomplete, the limited ability to generalize, poor reproducibility, reliance on tools, and data leakage that occurs during evaluation. The The paper examines the four studies and suggests a possible workflow for using AI responsibly in scientific research. Overall, AI can be It is a useful research assistant, but its suggestions have to be checked by way of transparent analysis and independent validation.

DOI: https://doi.org/10.5281/zenodo.23565050

Published by: vikaspatanker

Authors: Pratik Shankarrao Shirke, Professor A. P. Bangar, Professor S. B. Bhosale, Professor V. A. Karad

Abstract: Chronic and life-threatening illnesses are frequently found late, and assessing risk by hand is slow and uneven across large groups of patients. This paper offers a focused narrative review that brings together five studies on machine-learning-based disease prediction published by IEEE and Elsevier: a convolutional-network risk model built on regional hospital data; a hybrid random-forest and linear model for heart disease; a preprocessing-and-ensemble framework for diabetes; a smart monitoring system that combines sensor data and medical records through ensemble deep learning; and a study of breast tumour classification. The review follows three connected themes: the data sources and prediction models used, how strong and how comparable the reported evidence is, and the conditions that matter when prediction is carried into a clinical decision-support pipeline. Only one study could be read in full. It reports an area under the ROC curve (AUC) of 0.950, with sensitivity 0.789 and specificity 0.934, for a weighted ensemble on the PIMA Indians Diabetes dataset. The other four were assessed from their published abstracts and report, among other results, accuracies of 94.8%, 88.7% and 98.5%; these figures involve different diseases, data and metrics, so they cannot be ranked against one another. The review contributes a comparative table, a conceptual architecture for the proposed disease prediction system (not experimentally validated), and a prioritised agenda of research gaps.

DOI: https://doi.org/10.5281/zenodo.23564518

Published by: vikaspatanker

Authors: Pankaj Rajesh Bhakde, Professor Abhishek Popat Bangar, Professor S. B. Bhosale, Professor V. A. Karad

Abstract: IoT deployments bring together a wide mix of sensors, gateways, embedded devices, and software services. Finding the cause of a fault, or setting up a device the technician has never seen before, calls for knowledge that is typically spread across datasheets, manuals, incident tickets, and past reports. This review looks at retrieval-augmented generation (RAG) and large language model (LLM) techniques for IoT troubleshooting, using five selected studies that cover incident diagnosis and reporting, technical-data extraction, agentic answer validation, an industrial troubleshooting assistant, and conversational machine monitoring. Across these works, retrieval anchors generated answers to outside evidence, structured extraction makes technical specifications easier to query, and validation steps cut down on mix-ups between similarly named fields and on answers the evidence does not support. The reported evaluations show measurable improvements in technical question answering and task performance, yet they also expose weaknesses with hard questions, varied document layouts, source reliability, and generalisation to new deployments. We compare the methods and contributions of the studies, point out research gaps, and outline an evidence-centred architecture that joins document ingestion, structured device knowledge, hybrid retrieval, sensor-context reasoning, answer verification, and technician feedback. Our main conclusion is that a dependable troubleshooting chatbot cannot rest on fluent text generation alone: its advice has to be traceable to device-specific evidence and bounded by technical rules. Since the review draws on five chosen papers rather than an exhaustive database search, it is offered as a focused narrative review, not a systematic one.

DOI: https://doi.org/10.5281/zenodo.23564170

Published by: vikaspatanker