Host: The Japan Society for Management Information
A decision tree model is a useful representative classification model. Its effectiveness is shown in many previous papers for learning or prediction problems and a variety of decision tree generating algorithms have been proposed.Many of them can be considered to be the methods of that select one model from a possible tree model class by the given training data.Then the Bayes optimal prediction algorithm using the mixture model of all the decision tree models has been proposed.However, the objective variables were limited to be in discrete values in the method.In this study, we propose the Bayes optimal prediction algorithm that is able to deal with a continuous value.Then the proposed method is applied to a case study and its effectiveness is shown.