Research And Implementation Of WAP Log Mining

With the development of wireless data business from "barbarism period" to "cultivation period" and the keeping effort of Telecom Operators to regulate the telecommunication markets, precision marketing has become more and more important way to raise profits for service providers. We can get a great amount of potential data rules from data mining technology. Those data can reflect the trends in product operations, and offer an objective analysis basis for the marketing, which will greatly strengthen the production sales direction and increase profits.WAP log mining based on logging description, realizing the process from data extraction to marketing advices. The whole process includes the following steps: The ETL process of WAP log, the determination of the mining purpose, the choice of marketing variables, the foundation of the mining model, the analysis of the mining results , and the marketing program development.The WAP log records the user behavior on WAP pages. Since its complex format, it is difficult for sales person to analyze directly. Therefore the marketing measurements are been introduced as an interface between the technical person and sales person. Technology person extract data based on functionality and sales people concentrate on the precise marketing models established on those data.In the process of building precision marketing model, we use the following data mining technology: Clustering, Decision trees, Time series, Linear Regression Analysis, Correlation analysis and Neural Network.Compared with other algorithms, the use of clustering algorithm achieves the relatively objective classification for different users and areas. Using decision tree and the method of clustering to analysis will greatly improve the readability of the classification results. Using time series and linear regression analysis to forecast the product key indicators will provide an important basis for evaluating the server bearing capacity. The use of neural network can get the behaviors of active users, and then predict the potential users by the result of this.This thesis construct a framework for making differential marketing plans by regional characteristics, forecasting the key indicators of product, classifying the users , and finding potential users. It will dramatically enhance the service provider's understanding and control towards different users and explore a better path to precision marketing.