Classification and feature extraction of criminal occurrence points using CAEP with transductive clustering

In this study, we analyze bag-snatching in Fushimi-ku and aim to improve the author's previous study (Takizawa et al., 2010). The quantity of natural surveillance from a building is calculated by considering the wall as a unit. The number of pedestrians on the street is estimated by the random-walk method. Some new attributes are added, and discriminant analysis is performed by CAEP. We optimize the area of criminal occurrence class separately to improve the classification accuracy of CAEP. We use a method based on the concept of transductive clustering. We then propose a relationship between street crime and micro space.