Soil classification, crop prediction, and disease detection using ML and DL–“agro insights”

Abstract

India, renowned for its rich agricultural heritage, is ranked among the three world crop suppliers. Farmers face numerous challenges, including difficulty in selecting profitable crops suited to their soil and unpredictable weather conditions that affect yield predictions. To address these issues, various analytical methods have been employed in agricultural yield-prediction studies. Plant diseases are prevalent in agriculture, prompting the need for effective detection methods. Therefore, in this study, the proposed agro insights’ model aimed at assisting farmers in predicting or deciding the type of soil and crop to sow, which is implemented through ML and DL methods to predict the optimal crop to be cultivated by deciding diverse input variables such as the region, soil, and crop type. The accuracy of soil classification and crop recommendation is 93.3% using random forest technique and crop disease detection is 96% using CNN technique.