Recommendation Methods And A Preliminary Implementation Of Recommendation System For Mobile Tariff Packages

With the reorganization of telecommunication market, implementation of entire service operation and release of 3G license, profound changes has been taken for Chinese telecommunication market. Market competition turns to be intense, and the telecommunication companies are urgently promoting their service level and increasing customer's satisfaction from various perspectives. The mobile tariff package is an important tool for telecommunication companies to develop new customers, retain old customers and improve customer value. It is an urgent problem to be solved that how to use this tool to satisfy customers and make profit for companies. Therefore, it is particularly important to find effective recommendation methods and develop reasonable recommendation system for tariff packages from the perspective of customer requirements.Under this background, this thesis mainly studied the problem that how to provide users with humanized and accurate mobile tariff packages under the premise of diversification of consumer purchase behavior. This thesis put forward three main types of recommendation method for mobile tariff packages for different scenarios.Firstly, for the users whose occupation is the main influencing factor, this thesis proposed a recommendation method based on data classification techniques.This method can recommend the higher-weight package depending on the types of the users.Minimal neighborhood method is used when the data type is not determined during the classification process. A recommendation interface is designed, in which the occupation and income of users is taken as input conditions. The mobile tariff packages with higher weights is recommended according to user's input data type.Secondly, for the users who have historical consume data and want to know their own consumption habits, this thesis designed a recommendation interface, which applies quadric smoothing forecasting model and can obtain the history data of telecommunication consumption from the mobile phone number input by user, and then recommends mobile tariff packages based on the forecasted customer consumption data.Thirdly, for the users who can't estimate their consumption data accurately, this thesis proposed the following two recommendation methods:(1)A recommendation method based on the input of interval fuzzy numbers is developed. A recommendation interface is designed for various mobile tariff package properties including the minimum consumption, the maximum consumption and other consumer attributes semantics. A illustrative case is used to explain the recommendation process.(2) A recommendation method is proposed for customer input with triangular fuzzy numbers. A recommendation in which the users can input their demands in triangular fuzzy numbers is developed. Various kinds of membership functions of fuzzy item costs after fuzzy operations are considered. Nine commonly-used fuzzy sort methods are used to rank the mobile tariff package based on the final shape of the membership function of the fuzzy number.Finally, based on the abovementioned recommendation methods, a recommendation system for mobile tariff packages is developed under the development environment of Flash Builder and Visual Studio 2005 based on the database of SQL Server. The recommendation system consists of several functional modules including recommendation module, ranking list module and mobile tariff information model.