Comparative Study of Ensemble Models for the Prediction of Crop Yield

Abstract

Crop production is one of the most important sources of the Indian economy and one of the income sources for farmers in India. Prediction of crop yield in advance can guide farmers and improve revenue. In Agriculture Section, Maharashtra is one of the leading states from India. One of the principal foods in Maharashtra is Soybean. The proposed method uses a different ensemble model for prediction of Soybean crop yield of Marathwada region from Maharashtra. The crop yield data have been collected from the Indian Government's records. The environmental data like temperature details, rainfall details, and rainy days for the years 2001–2019 have been collected from India Meteorological Department. The dataset of crop yield and dataset of features have been preprocessed and different dominant features are identified using recursive elimination method. Different types of ensemble models have been applied and performance evaluation has been done using measures like mean absolute error, root mean squared error, and mean squared error. After analyzing performance of different ensemble models, we have found AdaBoost Decision Tree Regressor and AdaBoost Linear Regressor performed well as compared to AdaBoost Random Forest Regressor.