A. P. Shah Institute of Technology

The Indian agricultural sector is crucial for export revenue, and rural employment. However, current digital technologies limit holistic decision support, and fail to consider the combined effect of agronomic, climatic, and economic factors on farm productivity. These factors especially impact small and marginal farmers, whose agricultural productivity is highly affected by both climate variability and economic uncertainty. Therefore, the objective of this research was to create a new, holistic, and multilingual decision support system, which would provide farmers with real-time, multi-stage and contextualized recommendations throughout their entire farming lifecycle. Unlike other decision support systems, this proposed system utilizes gradient boosting ensemble machine-learning models to evaluate and predict crop suitability using a combination of soil nutrient profiles and environmental information. The model used to collect and validate agronomic input data uses a mediator-assisted data collection methodology, where trained mediators employed by government or non-governmental organizations assist in collecting and validating agronomic input data, resulting in improved data quality and increased access to decision support tools for small and marginalized farmers. The system incorporates a fertilizer prediction module that leverages soil characteristics, nutrient profiles, and crop-specific nutrient requirements to enable precise and efficient fertilizer usage. In addition, a short-term weather forecasting component based on deep learning models such as GRUs and RNNs to identify temporal patterns in historical weather data and generate early warnings for weather-induced risks. Experimental evaluation shows that the system offers reliable predictions and effective decision support, confirming the practicality of a holistic, multilingual, and scalable agricultural decision support system capable of fostering resilience, inclusivity, and sustainability.

In a special ecosystem, such as the mangrove region of the Sundarbans belts, conventional field-based surveys are of limited use due to less frequent observation cycles and restricted on-ground coverage. Monitoring of ecologically sensitive regions for forest degradation requires an approach that can capture gradual structural changes over a long period of time. The studies offers an analytical framework by using satellite data that evaluates the forest conditions using multi-temporal vegetative indices. By examining various indices across a span of years, the framework is able to identify gradual degradation trends that may not be evident through single-year data analysis. The methodology of the study focuses on temporal variation in Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), Enhanced Vegetation Index (EVI), and Soil-Adjusted Vegetation Index (SAVI) to examine various parameters such as moisture dynamics, canopy vigor, and soil exposure patterns. The application of the proposed system approaches to the Sundarban region between the years 2020 and 2025. This reveals a consistent downward shift across all the indices, indicating ecological stress and reduction in vegetation density for a long duration. The results demonstrate that the multi-temporal index analysis offers a scalable and repeatable mechanism for long term forest monitoring, supporting data-driven environmental assessment and policy planning by the authorities.

Natural Language Processing (NLP) is largely based on text classification as its major working process for automatic analysis in different domains. The discipline of text classification has opinion polarity analysis as its significant subset that examines the sentiment in written content. The study explores machine learning methods for examining the feelings of English writers' English literary reviews in Indian writers and determines the style of classification between positive, negative and neutral criticisms. The study used usual machine learning classifications Naive Bayes (NB), Support Vector Machine (SVM), Random Forest (RF) and sentimental analysis. The model performance is improved by applying TF-IDF and word encodings to extract features. The corpus consists of curated reviews collected from various online literary sites as well as book reviews and social media platforms. The test of the model determines its effectiveness using standards such as accuracy, accuracy, recall and F1-score. Among the evaluated models, SVM achieved the highest accuracy of 87.5 %, followed by NB with 86.9 %, while RF recorded the lowest accuracy of 84.2 %. These results promises to deliver the outstanding performance of SVM for sentiment classification of English literary reviews. The study advances automated literary analysis algorithms that identify reader emotions and offer suggestions. It also produce insights that examine how Indian authors are received internationally through sentiment analysis employing computational methods, the project combines literary analysis methodologies with machine learning.

VayuNetra is an artificial intelligence platform that monitors, analyzes, and reduces carbon emissions, especially transport carbon emissions, by helping communities monitor, analyze and reduce them. VayuNetra uses rule-based models to estimate CO2 emissions and natural language processing and location services to help users estimate their carbon footprint and provide environmentally friendly tips. The platform promotes sustainable living by providing a digital carbon credit platform, role-based access and interactive dashboards to track progress at individual and community level, integrating real-time environmental data and AI-based pollution peak predictions to enable users to make informed decisions and minimize their environmental impact.

Background: Type 2 Diabetes Mellitus (T2DM) is one of the most prevalent chronic metabolic disorders worldwide and is associated with severe health complications and increased healthcare burden. Early identification of individuals at risk is essential for timely intervention and disease management. Recent advances in artificial intelligence (AI) and machine learning (ML) have provided promising approaches for improving disease prediction and clinical decision-making. Objective: This study aimed to develop and evaluate machine learning models for the early prediction of Type 2 Diabetes Mellitus using clinical and lifestyle risk factors and to identify the most influential predictors associated with diabetes risk. Methods: A publicly available diabetes dataset comprising 768 participant records, including 268 diabetic and 500 non-diabetic individuals, was utilized. Data preprocessing involved missing value handling, normalization, label encoding, and outlier detection. Multiple supervised machine learning algorithms, including Logistic Regression, Decision Tree, K-Nearest Neighbor, Support Vector Machine, Random Forest, and XGBoost, were developed and evaluated. Model performance was assessed using accuracy, precision, recall, F1-score, ROCAUC, and confusion matrix analysis. Explainable Artificial Intelligence (XAI) was implemented using SHapley Additive exPlanations (SHAP) to determine feature importance. Results: Among the evaluated models, XGBoost demonstrated superior predictive performance, achieving an accuracy of 93.4%, precision of 92.1%, recall of 91.5%, F1-score of 91.8%, and ROC-AUC of 0.97. SHAP analysis identified blood glucose level, body mass index, age, and family history of diabetes as the most significant predictors of diabetes risk. The findings indicated that ensemble learning approaches outperformed conventional machine learning algorithms in diabetes prediction. Conclusion: The proposed AI-based framework effectively predicted Type 2 Diabetes Mellitus using a combination of clinical and lifestyle risk factors. The integration of machine learning and explainable artificial intelligence provided both high predictive accuracy and model interpretability, highlighting its potential application as a clinical decision-support tool for early diabetes risk assessment and preventive healthcare interventions.