蜂窝网络中用户流量使用模式挖掘

Based on the traffic data of more than 300,000 users collected from a carrier\'s cellular network, more than 30,000 heavy users who have significant influence on the network are extracted first, and then feature vectors that measure user preferences for different time periods are created to explore the time rule of the traffic usage of the heavy users. Then, the appropriate clustering algorithm is selected, and the three evaluation indexes of the elbow method, the contour coefficient and the CH index are used to obtain the four flow usage patterns of heavy users. Finally, the main characteristics of each user are analyzed, which provides reference suggestions for operators\' network optimization and precision marketing.