Artificial neural network is a common method which has been used in many works. This paper presents a comparison of prediction methods for alum dosage using in water supply treatment process. In this research, we compared results from M5P, M5Rules and REPTree to the results from multilayer perceptron, one type of artificial neural network. Six input variables, i.e. turbidity, alkalinity, pH, conductivity, color and suspended solids relating to reaction of coagulation were used. The data in this research had been collected from Bangkhen Branch Office of Metropolitan Waterworks Authority, Thailand from 1 January 2006 – 31 July 2015. Our experimental results showed that the M5Rules method yielded the highest accuracy to predict Alum dosage comparing to other methods run in this study. For M5P and M5Rules, building model by using smoothing procedure and unpruned technique appears to give out the best model to predict with the highest accuracy.