Predictive modelling of rice productivity in Tamil Nadu: Insights from 10 years of agricultural data

Rice is a vital crop in Tamil Nadu, India contributing significantly to food security and rural livelihoods. This study develops a predictive model for rice productivity using agricultural data from 2015 to 2024, encompassing area cultivated, production volume, and yield across multiple districts. The objective is to forecast rice yield for the next five years (2025–2029) using statistical and machine learning techniques. The dataset was cleaned and analysed using multiple linear regression (MLR), time series forecasting (ARIMA- Auto-Regressive Integrated Moving Average), and random forest algorithms. Among these, the random forest model achieved the highest accuracy, with an R² value of 0.92, RMSE of 185.6 kg/ha, and MAE of 124.3 kg/ha. Key predictors included seasonal rainfall, sowing window and fertilizer application rates. The model revealed a consistent upward trend in productivity in districts adopting improved agronomic practices, with Chengalpattu showing a 12.4% increase in average yield over the decade. Forecasts for 2025–2029 suggest a potential yield increase of 6–9%, assuming stable climatic conditions and continued adoption of best practices. These insights can support agricultural planning, optimize input use and guide policy interventions for sustainable rice production. This research demonstrates the value of data-driven modelling in agriculture and provides a scalable framework for yield forecasting. Future work will integrate satellite imagery and farmer-level survey data to enhance spatial resolution and model precision.