MandiMitra: A Phenology-Aligned Computational Intelligence Framework for Spatiotemporal Crop Risk Assessment and Market Price Forecasting

Smallholder farmers in rain-fed regions must decide what to sow and when to sell under increasing climate variability and mandi price volatility.MandiMitra is an integrated mobile decision-support system that couples a phenology-aligned crop risk model with a short-term market intelligence engine.In Module 1, a user-defined sowing date is mapped to crop growth stages; daily rainfall and derived extreme-event features (e.g., harvest wetness, flowering dry spells) are used to train an interpretable Random Forest classifier that outputs stagewise risk labels and human-readable reasons.In Module 2, daily modal prices from Agmarknet are transformed into lag, rolling-statistics, and seasonal features and fed to an XGBoost regressor with walk-forward validation to forecast 15-day prices and highlight favourable selling windows using seasonal indices.Experiments on Maharashtra data (2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018)(2019)(2020)(2021)(2022)(2023)(2024) show that the price engine achieves 3.8% MAPE for soybean and reduces RMSE compared with an ARIMA baseline, while the risk model attains 82% accuracy and 0.88 recall for detecting high-risk soybean years in Pune.By converting raw weather and market data into actionable advisories, MandiMitra supports timely sowing and profitable sales decisions.