Crop Yield Prediction in India: A Comparative Analysis of Ensemble Techniques

This study focuses on solving several problems associated with Indian agriculture, which is a lifeline for 64% population in India. The sector is very wide with vast arable land, yet, the output remains low rendering a lot of challenges among the farmers and their contribution to the national economy is equally dismal! The proposed course of action involves employing advanced scientific techniques such as artificial intelligence, machine learning, and data modeling that will provide accurate estimates on harvests. This review article considers various methods and algorithms in agronomy including machine learning-ensemble approach for crop disease detection and yield estimation in precision farming. This research has used a Kaggle dataset which includes soil minerals, temperature, moisture, rainfall, crop costing, and others, in order that the datasets are cleaned toward achieving relevance to accuracy using the data pre-processing technique. The methodologies section introduces three distinct approaches: The Stacked model using the following models as base: Random Forest Bagging, AdaBoost Decision Tree, AdaBoost Random Forest, and Logistic Regression as Meta-model (Mangla, Shinde, Mehta, Sharma, Mohanty eds 2022 Handbook of Research on Machine Learning: Foundations and Applications. CRC Press). They are a group of procedures which enable prediction of crop yield performance by different predictors. This means that each of the group techniques performed well and stack being the best way to predict a yield for this crop. Evaluation indices like accuracy, precision, and the recall almost inescapably support Stacking that is highly capable of improving the correctness and reliability in production. Lastly, these results point out to the significance of ensemble learning methods for improving the predictabilities of crop yield (Sharma et al. in Comput Electr Eng 96, 2021;Mangla et al. in Innovations Syst Softw Eng 18:301–308, 2022). Stack is one of the best options in such cases as it displays highest accuracies in the market. Thus, such kind of models will provide better forecasts, which enables farmers and other stakeholders to allocate inputs or crop varieties.