Adaptive boosting of weak regressors for forecasting of crop production considering climatic variability: An empirical assessment

Abstract

Crop yield forecasting based on different climatic conditions for coastal regions is a critical process. In this study, regression based adaptive boosting prediction model is presented, using the datasets of Kharif and Rabi seasons along with the climatic features of three coastal districts belonging to Odisha located in India. This study discusses and experiments on the different weak regressors, such as: linear, lasso, ridge, SVR regression, proposes strong predictors by avoiding the shortcomings of individual weak regressors and propagating the benefits of AdaBoost to improve the predictive accuracy on learning problems. AdaBoost helps to get a combined output of the weak regressors into a weighted sum that represents the final output of the boosted strong regressor and also the output of the weak regressors which is likely to be twisted in favour of wrongly predicted instances adaptively. It has been observed from the experiments that, the decision of weak regressors vary due to frequent, inherent attributes of climatic conditions for crop production. Obtained numerical simulation results in terms of errors, various performance measures and statistical analysis demonstrated have highlighted the attractiveness of the proposed strong regressors compared to weak regressors forecasting methods for crop production.

Keywords

• AdaBoost
• Linear regression
• Lasso regression, ridge regression
• SVR linear regression
• SVR polynomial regression, SVR RBF regression
• Crop yield forecasting