Research And Implementation Of Traffic Data Collection And Analysis Platform Based On Android

With the rapid development of3G wireless technologies and the expanding coverage of4G networks, the explosive growth of mobile data services has arisen. On one hand, the growth of data services will bring more revenue and opportunities to the operators. On the other hand, the growth will make it more difficult for the operators to configure and optimize the networks. Therefore, it is of great significance for wireless network planning and optimization to collect, analyze and predict the mobile Internet traffic.As all kinds of wireless technologies emerge, users get more and more frequent access to the Internet via handheld devices, especially those using Android, an open source operating system with a very high market share. A wide variety of Android devices have covered all customer groups, which makes Android a typical system to study. In this thesis, a traffic data collection and analysis platform based on android is designed. The platform consists of a terminal-side data collection system based on Android and a server-side data analysis system. Firstly, the overall and module design of the terminal-side data collection subsystem is described in detail. The terminal-side system can collect various types of service data of cellular subscribers, such as voice service data, short message service data and traffic data. The GPS coordinates and cellular signal strength when the traffic is generated are also recorded. Then, the overall and module design of the server-side data analysis subsystem is presented. Due to the large amount of traffic data caused by the increasing number of users and the long-lasting data collection process, a single server may not accomplish the computing tasks. To deal with this problem, a big data processing module based on Hadoop is designed for the server-side system. At last, the prediction of wireless network traffic based on SVR model is performed by using real traffic data collected by the platform. It is shown that the precision of SVR-based prediction is close to90%in different time scales. At last, the prediction of wireless network traffic based on SVR model is performed by using real traffic data collected by the platform. The precision of the SVR-based prediction results is close to90%in different time scales, which proves SVR an effective method to predict wireless network traffic.The platform we design in this thesis can provide an integrated solution for cellular service data collection, analysis and prediction. It also works well with big data.