Mining Resident Activity Pattern Based on Spatio-temporal Trajectory Data

The research on residents' activity pattern based on spatio-temporal trajectory data is helpful to the optimization of urban operation. At present, one of the difficulties in the research is how to determine the regularity of residents' activities when the labeled samples are sparse. We propose an improved periodic decision algorithm of sliding window, combined with feedforward neural network to search outlier activity patterns. Experimental results show that our method can effectively classify the overall travel features and quantify individual activity abnormalities.