Smart Agriculture: Comparative Analysis of ML Models for Crop Yield Prediction

Agriculture remains a widespread yet low-income occupation in India. As a cornerstone of civilization, it plays a crucial role in the country’s development. Given that India is largely an agrarian nation with an economy heavily dependent on crop production, it can be argued that agriculture has the potential to sustain and drive the nation’s economic growth. A comprehensive literature review has been conducted on various machine learning and deep learning techniques for crop yield forecasting, including MAWT SVM, IOFLSTM, Conv LSTM ViT, and FMIG RFE SVM, among others, highlighting their strengths and limitations. While these models have shown effectiveness in increasing prediction accuracy, they frequently encounter challenges, including high computational demands, overfitting, restricted feature utilization, and reduced generalization across varying environmental conditions. This article utilizes multiple datasets to predict crop yields using ML models and then compares their performance. Also analyzes the relationship between predicted and actual crop yields. The DT gives high prediction accuracy across different conditions.