Since Global Positioning System (GPS) modules have been attached to smart phones, many researches focus on how to obtain the personal trip (PT) information from the mobile phone.One of the challenges is to infer trip purpose (or identify activity type) from these continuous GPS data.This paper focuses on obtaining trip purpose with the methodology of Support Vector Machines (SVM) which has not been utilized in this field in the existing researches.Three different kinds of SVM are applied and one-against-rest SVM shows a higher accuracy.Furthermore, attributes related to activity intensively influence the accuracy of trip purpose inference.Also, excluding "others" purpose from the trip purpose list improves the inference accuracy.The final finding is that treating continuous business trips as a single business trip will not always improve the accuracy but demonstrates the significant influence simultaneously from at tributes related to activity, demographic and trip.