AgriSmart: An Explainable Multi-Criteria Crop Recommendation System using Simulated Real- Time Data

Abstract: Crop selection in Indian agriculture is largely governed by traditional knowledge and guesswork, leading to yield mismatches, nutrient imbalance, and financial risk. This paper presents AgriSmart, a cloud-enabled, explainable, multi-criteria crop recommendation system that integrates a Random Forest classifier, the TOPSIS multi-criteria decision-making method, and SHAP (SHapley Additive exPlanations) to recommend and rank suitable crops based on soil nutrient levels (N, P, K), pH, and climatic parameters (temperature, humidity, rainfall). Unlike single-output, black-box crop advisory tools, AgriSmart ranks the top six candidate crops on a 0–100% suitability score, explains the contribution of each input feature to the recommendation, and generates a downloadable soil health report with fertilizer and remediation guidance. The system is implemented as a responsive React.js web application that communicate s with a cloud-hosted machine learning backend, requiring no local server infrastructure. Experimental evaluation shows that the integrated pipeline produces consistent, transparent, and actionable recommendations within a few seconds, indicating strong potential as a decision-support tool for farmers, agricultural officers, and researchers.

Keywords: Crop Recommendation, Random Forest, TOPSIS, Explainable AI, SHAP, Precision Agriculture, Multi-Criteria Decision Making, Soil Health