Pearl Millet Yield Prediction: A Comparative Analysis for Forecasting Models

The crop yield estimation is an important aspect in agri-supply chain and forms the basis of decision-making for its different functional areas. The crop yield has a direct impact on farmer income, logistics planning and its infrastructure development. In the current study, pearl millet yield in Indian context is predicted considering available secondary data (rainfall, year, temperature, soil moisture, and pressure) for six districts (Nagaur, Jodhpur, Alwar, Barmer, Churu, and Jaipur) of Rajasthan from 1966 to 2020. The pearl millet yield is predicted using various predictive analytics tools such as multivariate adaptive regression splines (MARS), random forest (RF), multiple regression, TreeNet, and classification and regression tree (CART) in Minitab version 21. The root mean square error (RMSE), correlation factor (R), and adjusted R-squared were the three accuracy measures utilized to validate the models. All models were statistically tested using out-of-bag or with fivefold cross-validation. The results show that fairly accurate prediction was obtained The results of the study indicate that the TreeNet model outperformed other predictive models, achieving an R-squared value of 74.59%, demonstrating its superior capability in predicting pearl millet yield. Temperature has the highest relative importance based on the predicted model, followed by pressure, rainfall, and soil moisture. The present work will be a valuable addition to the pearl millet yield for casting literature in specific.