Research Of The User Mobility Prediction Model In Mobile Internet Based On Hadoop Platform

With the continuous development of the mobile network, the number of the mobile users has shown a rapid growth. More and more people choose to use the mobile terminals,such as the mobile phones and tablet computers, to get the services they need. This process will produce large amounts of mobile communication data. By analyzing these mobility data,we can learn user's movement patterns and mobility behaviors, which can help us implement the mobility prediction of the users in the Mobile Internet. Researching the mobility prediction model is an important topic which can help us understand the human behavior, social connections or personal preferences. The researching results will help us solve the practical problems in our daily life, for example, the spread of human diseases and the application recommendation in mobile communications.It plays an important role and produces positive influences on the promotion of the mobile communication technologies.The data produced during the process of mobile communication are heterogeneous and mass. The traditional data processing technology can not be able to meet our requirements of data storage and analysis.Therefore, we choose to use the Hadoop distributed platform to conduct the mobility data analysis and prediction.Firstly, we introduce the architecture of the Hadoop software, with the operation principle of Hadoop conducting the data storage and data computing. Then, we build the mobility prediction model of mobile Internet users based on the Hadoop platform. Next, we research how to extract important places from the user trajectories. We implement different algorithms and evaluate the results by the Point of Interests (POI)identification indicator. Then, we analyze the mobility data to learn the predictability of individual location .We propose a new method based on the instantaneous entropy and transition probability, which can discover the regular and irregular places in the user trajectory. Finally, we implement the mobility prediction of the mobile users. Meanwhile, we propose a practical model based on State Based Prediction (SBP) method which can address the shortcomings of the traditional mobility predictors and improve the prediction performance.