Predicting Consumers Purchase of Agricultural and Livestock Products Using Data Mining Techniques

Abstract

The purpose of this study is to compare the prediction power of agricultural product purchase analysis using the decision tree model and neural network model with the existing econometrics model. The research subjects are beef, Chinese cabbage, radish, red pepper, garlic and onion, which are very vulnerable in terms of supply and demand at the Korean agricultural products markets. Using the three models, we predicted the 1,314 households purchase of agricultural products with the 2016~2017 consumers panel data provided by the Korea Rural Development Administration and the Internet search index obtained from the Naver Data Lab. The main results of this study are as follows. First, based on the MAPE, the decision tree model had the highest predictive power, while the panel Tobit model had the lowest predictive power. Second, with the exception of some products, the predictive rates of peak season were higher than those of off-season. Therefore, the data mining technique is considered to be a complementary method to the existing econometric models in agri-food consumption analysis in terms of predictive power.