Machine learning technologies have transformed decision-making and data-processing methods, resulting in global economic development. This study investigates the possible benefits of incorporating machine learning techniques into modern agricultural systems. This paper provides a summary of the latest machine learning utilization in agriculture today, outlining both major challenges and opportunities. Forecasting crop yields is dependent on a wide range of interdependent elements, such as crop sowing and harvesting season, soil, temperature, humidity, rainfall, and management practices. The current research examines the performance of machine learning approaches for predicting crop output using a huge dataset of weather conditions, soil, season, area, and crop production parameters collected from 33 Indian states between 1966 and 2017. From the result analysis, it is inferred that the CatBoost regressor combined with SHAP and CatBoost Regressor models’ performance was slightly poor for training dataset. According to the results of the other models, the Random Forest Regressor, out of the six machine learning models, had the highest accuracy during the training phase. During the validation stage, models such as CatBoost Regressor with an average R2 values varied between 0.967 and 0.866, while CatBoost Regressor + SHAP, where averaged R2 value ranged from 0.934 to 0.9171, produced the most accurate findings. While SVR performed better in the training phase, RF yielded more accurate results in other combinations. The RMSE is reduced to average yield with optimal crop yield and production data.