Design And Implementation Of A Specific People Classification Model For Spatio-temporal Data

The use of spatio-temporal data to analyze the status quo of crimes and predict the trends in crime has always played an important role in the investigation of specific people such as criminals by the public security departments.With the development of information technology,through sensors like mobile terminal and monitoring equipment,acquisition of people's mobile spatio-temporal data,from which to identify specific people such as criminals,has become the focus of public security departments in recent years.With the explosive growth of collected data,the traditional way of using GIS for investigation has been stretched.In this thesis,according to the spatio-temporal data of mobile terminal networking collected by the public security department of a certain city,and the MAC addresses of the specific people's mobile terminal provided by the public security department,a specific people classification model is designed and implemented.The main work of this thesis is as follows:First,clean the original collected data in a variety of forms,extract the spatio-temporal data fields of interest to this thesis.Second,in order to reduce the high dimensionality of the original data,this thesis proposes two kinds of spatio-temporal feature extraction algorithms from different perspectives.The TSG algorithm divides the time and the space uniformly and takes the statistical data after the spatio-temporal information division as the features.The STPS algorithm takes the proportion of spatio-temporal information and the similarity of daily activities as the features,from the perspective of work-day schedule of the office workers.These two algorithms greatly reduce the feature dimension of the original data set,and at the same time,make up for the problems of repeated or missed data acquisition of some AP devices.Third,continuously improve the performance of this classification model in this thesis through two-stage ensemble learning.In the first stage,according to the performance indicators concerned in this thesis,we identify the classification algorithm that is suitable for the training set generated by the TSG and STPS algorithms,and then apply the Bagging algorithm in the ensemble learning field to the classification models,to obtain a classifier with higher performance.In the second stage,using the generalized ensemble learning method,the TSG classification model and the STPS classification model are combined according to the specified rules to further improve the performance of the classifier.The experimental results show that the specific people classification model implemented in this thesis is effective in assisting the public security department to identify the specific people.The classification model implemented in this thesis has been applied to the person analysis system of public security department in a city.