Infrequent Data Binning Method for Improving Prediction Performance: The Case of Rice Productivity Prediction model

This study investigates ways to improve the performance of rice productivity prediction model by employing the infrequent data binning method. Binning in this study is a technique to reassign infrequent data outside a specific scope, back in to the boundary value of the scope. The main findings of this study include: first, the binning method based on reassigning infrequent data contributes to improving the prediction performance of the model in question. Second, the effects of improvement differ depending on the length of the tail of a distribution. Third, there are no interaction effects due to combination of binned variables involving in different distribution categories with different length of long-tail