Ever since global positioning system (GPS) modules have been attached to smart phones, much research has focused on how to obtain personal trip (PT) information from them. One of the challenges is identifying activity type (or inferring the purpose of the trip) from these continuous GPS data. This paper focuses on obtaining the type of activity using several machine learning methods and comparing the results. The comparison is implemented from the perspective of accuracy and time cost in the phases of data training and prediction. After applying four machine learning methods to the data set obtained from 30 individuals in Nagoya, Japan, a classification tree method demonstrates superiority over support vector machine (SVM), neural network (NN), and discriminant analysis methods.