Using four data mining techniques to predict grain yield response of winter wheat under organic farming system

Abstract

Organic farming is one of the resource-conserving and environmentally friendly systems that achieve the sustainability principles. An essential issue for sustainable agricultural planning is the accurate yield estimation for the crops involved in the crop rotation. In this study, the potential of predicting grain yield for organic winter wheat under varying soil and climate conditions was conducted by applying four different data mining techniques: multi linear regression (MLR), general linear model (GLM), artificial neural networks (ANN), and regression trees (RT). Considering the modelling accuracy and prediction accuracy, RT is the most robust technique for predicting grain yield of winter wheat at the study sites. MLR and GLM produced the poorest results for the data sets compared in this research. Such poor performance might be due to insufficient MLR and GLM techniques to model non-linear regression present in complex soil-weather-land management interactions. The ANN model ANN is also a promising tool for predicting grain yield of winter wheat particularly under low sample size, however, optimum model structures require further attention.