Krishi-Stats: A Web-based System for Crop Price Prediction using Machine Learning Approach

Abstract

Agriculture is the main livelihood in India. Most of the people earn bread and butter through farming, but the farmers are not getting enough profit and the field is facing growth downward due to irregular rainfall, high volatility in agriculture commodity prices and uncertainties in production. The objective of this study is to design and implement an automated crop price prediction system with best suitable machine learning technique, as well as displaying prediction results on website Krishi-Stats designed for easy understanding for Farmers. In this study, three machine-learning (ML) algorithms, ARIMA, VAR and XGBoost are applied on large historical data collected from government website. The ML algorithms compared with their root mean square error values (RMSE). As XGBoost has given optimum RMSE value of 0.94, has been selected as the prediction system engine of our website Krishi-Stats. On website, the crop prediction prices are plotted for all twelve selected crops and visualized using prediction graphs.