With the rapid development and popularization of WiFi technology, indoor positioning technology based on WiFi has become a hot spot. At present, research on WiFi is mainly focused on two-dimensional positioning in the field of indoor positioning, which is less concerned about the 3D positioning. In this paper, a method of WiFi indoor 2.5D positioning based on support vector machine (SVM) is proposed. The method first uses the WiFi signal of each floor to identify the floor by SVM, and which obtains the position along the altitude direction. Then, the weighted k-nearest neighbor (WKNN) algorithm is applied for plane location estimation over a set of WiFi location fingerprints. Thus, indoor three-dimensional positioning is realized. Through improvement of the AP selection method, the accuracy of plane positioning is enhanced. The experimental results show that the recognition accuracy rate of the floor is 99.09%, and the average error of the indoor location estimation is 0.63m.