The application of the cart and chiad algorithms in sugar beet yield prediction

Abstract

Yield prediction is a very important agricultural problem. Any farmer would like to know, as soon as possible, how much yield he can expect. The problem of predicting yield production can be solved by employing data mining techniques. This study evaluated the feasibility to predict the yield at Khuzestan Province in Iran using CART and CHAID algorithms. The analyses were performed using IBM SPSS Modeler 14.2. Three cropping seasons from 125 farms were selected between 2015 and 2018. The most important attributes were selected and the average yield was classified according to a decision tree. The data was partitioned into training (70%) and testing (30%) samples. The decision tree, including nine independent variables and 29 nodes, was produced through CART method. The decision tree, including nine independent variables and 39 nodes, was produced through the CHAID method. The CART and CHAID algorithms were evaluated using linear correlation and mean absolute error (MAE). Maximum precision of model in training part relevant to CART algorithm was equal to 95%, in testing part relevant to CART algorithm was equal to 93%. According to models′ precision, the results showed that CHAID and CART models were stable and suitable for prediction of sugar beet yield.