Data Mining In The Data Traffic Management

Under the new situation of the mobile Internet, telecom operators need to make the shift from traffic operators to data traffic management, which accommodates the needs of future data traffic business development and solves the crisis of the traffic pipeline and quantity-price scissors problem. With data traffic scale upgrading there is a serious imbalance of the data traffic resources, for example:the56%of users of XX mobile operator in July2012have the data traffic, nearly50%of them monthly data traffic in5M below and they consume only1.44%of the resources.To solve the problem, the article is based on precision marketing which recommends suitable products to the right users at the right time. The article focuses on solving how to find the appropriate users. Decision tree model predicts the potential high-traffic users in the low-traffic users. Creating a decision tree model includes clearing the business objective, designing variables, data preparation, understanding data, building the model and model evaluation steps. After clearing the business problem, we design the wide table of modeling requires and implement the wide table through SQL. Then we choose smart variables and explore the relationship of the smart variables and target variable to determine the import of the decision tree model variables. We establish the model through training set and then score the test set by the model to assess the effect of the model. With the model we can predict45%of potential high-traffic users through10%users.In order to achieve better marketing results, the article proposes the establishment of a unified view of user data traffic behavior. The view should contain more information of the data traffic usage of users. The view can support us to analysis the right time and the right products to recommend to the candidate users.