Crop Prediction and Recommendation Using Ensemble of DL Models

Abstract

Agriculture remains the primary source of income in India and is characterised by a variety of crops, soil types and climatic conditions. This study suggests an additional ensemble model serving to give effective and speedy predictions and recommendations for crops. During the study data from nearly 8 distinct features was collected from various databases and 2201 instances were finalised. The data focussed on climatic conditions such as temperature, rainfall, crop type and soil features, particularly the ratio of nitrogen, potassium and levels of phosphorous. Research indicates that algorithms such as Neural Networks and XGBoost share high effectiveness and accuracy in developing crop yield prediction models. Extensive research conducted shows that the ensemble of XGBoost and MLP Classifier algorithms provide an accuracy of 99.39%. By predicting crop yield based on historical data, the study aims to give sound recommendations on the crops to be cultivated under various weather and soil conditions.