Machine Learning Enabled Crop Price Prediction : A Review

Abstract

The present study evaluates the application of machine learning algorithms for predicting agricultural product prices. Accurate crop price prediction is important for farmers. Earlier mathematical models have struggled to incorporate non-linear relationships in agricultural product prices, which are influenced by environmental and economic conditions. This has driven the growing development of machine learning and data mining techniques for crop price forecasting. Long Short Term Memory (LSTM) models in particular capture time-dependent patterns and lagged effects of environmental variables on price fluctuations, improving forecasting accuracy. SHapley Additive ExPlanations (SHAP) is also utilized to interpret feature importance, aiding transparent decision-making in agriculture. Furthermore, ensemble models and XGBoost have demonstrated robust performance, achieving high accuracy on standard datasets despite their potential challenges that remain in the data. Accessibility feature selection, model interpretability, and scalability for farmers in countries like India, where digital literacy and market awareness are limited, ML-based systems can offer actionable insights such as expected future prices, suitable sowing times, and storage decisions. This can help avoid premature crop sales and ensure better returns. The integration of environmental data, commodity prices, and advanced ML models enables proactive planning and enhances income stability for smallholder farmers as agriculture faces increasing uncertainty due to climate change and market volatility. The development of accurate, interpretable, and scalable price prediction models holds promise for sustainable agricultural growth and resilience.