Applying supervised learning to find out water user’s behavior through ability to pay for irrigation water fee

Abstract

Supervised machine learning is considered as one of the methods to find out variable relationships more informative compared to traditional statistical methods. In this article, both traditional statistical analysis and supervised machine learning approaches are used to study consumer behavior through their willingness to pay for irrigation services. 222 households in Nam Dinh, Thai Nguyen, and Phu Tho province were investigated. By the regression model, the results show that the variable area (DT) and yield of winter-spring crop (NS_DX) directly affect the household’s willingness to pay. The result shows that the majority willing to pay of households for irrigation water are higher than the current level. However, by the supervised machine learning approach, the errors of the model, predictions of the household’s payment level, and solutions to avoid overfitting are also shown, which could not be implemented if using traditional statistical analysis.