Value Discovery And Behavior Prediction Based On Location Data

Recent years, with the explosive growth of smart phone, almost everyone becomes an important producer of information. Taking China as an example, with over 800 million mobile phone users, there are about more than 11 billion minutes of talk, 1.7 billion short messages and 2.3 PB data traffic produced every day. What is interesting is that most of the occurrence of the data is closely related to the location of the users. Based on this background, researches based on data of location is developing rapidly and quite a lot of them have got excellent results.Based on these previous researches, we chose the location data of mobile users provided by mobile operators in a month and carried on some related researches. On account of the problems of the location data model in dealing with individual user, we proposed an improved method of modeling and analyzed this model comprehensively.The development of a city gradually fosters different functional regions, such as educational area, residential area, commercial area, etc.and between these regions there exists different social information due to human activities. A better understanding of these relationship will be very helpful for us to understand human diffusion laws and predict the spread of human beings. In this paper, we used Natural Language Processing(NLP) method to discover the relations between regions quantitatively and Region Activation Entropy Model (RAEM) was used in detecting the social relations hidden between the regions. By measuring the Region Activation Entropy (RAE) based on RAF, we find an 88% potential predictability in regional mobility and confirmed that human mobility pattern had a high degree of regularity between different regions.At last, given the fact that human mobility pattern is slightly affected by the relationship between regions, we created a Markov model for prediction of position added by Region Activation Force (RAF). In this way the accuracy of position prediction was improved by 23.5% compared to other methods.The experimental result shows that our research will be helpful to the construction of smart city in the regional planning, traffic warning, location recommendation and other projects.