Human Behavior Recognition And Similarity Analysis Based On Location Data

With the development of technology, we are able to capture and storemassive data. Due to mobile phone, we can record our daily location,communication and mobile information continuously. From these data, abridge is built to connect basic data with higher level behavior by extractinginformation of human behavior. The research of human behavior patternwhich combines knowledge of pattern recognition, data mining andrecognition science becomes a great force to these fields.Since the bright theoretical and application future in psychology,economics and sociology, many researchers have done a lot of work in suchfields. However, there is still a long way to go. In this paper, I studiedcurrent methodology in human behavior pattern recognition. And my workis mainly based on Reality Mining database conducted by MIT Media Lab.The contents of this paper include:Study traditional method, status and problems in human behaviorpattern recognition. Put out major obstacles in such fields and highlight theimportance and its bright future in real life.With deep study in Reality Mining database, I adopted the PrincipalComponent Analysis method in such field. Define Eigen behavior of humanbehavior and explain its real meaning in life by using correlationcoefficients. Then use these Eigen behaviors to reconstruct and predict asubject’s daily routine. Also, using MATLAB to visualize human dailyroutine and Eigen behavior.Successfully combine Eigen behavior with Euclidean distance toclassify different groups, individual identification. Construct subjects’social network and analyze its characteristic by using Pajek. From what we have done in my research, it shows that it’s practical tomining human behavior using cell phone location data. Also, with these data,we can develop more applications in human behavior prediction, smarttraffic and location based service.