Association Rules Mining Algorithm And Its Application On Telecommunication Industry

Recently, the technique of Data Mining is widely concerned by the international experts in the fields of Artificial Intelligence and Database. Mining association rules is one of the most important branches of it. At the same time, mining association rules in transaction databases is mostly applied to the merchandise fields. With the severe competition and market saturation in the telecommunication field, improving and exploring the value of existing customers have become the primary task in business analysis of operators. Cross-selling, as a new marketing method, which is based on the mining association rules, responds to this change perfectly.In this paper, characteristics and disadvantages of the existing Fuzzy Association Rules mining and Positive and Negative Association Rules mining are analyzed. Then a Novel Fuzzy Positive and Negative Association Rules(FPNAR) algorithm is proposed to deal these problems. At last, FPNAR algorithm is used for a practical example of telecommunication projects and feasibility of cross-selling in telecommunication industry is proved.The main contributions of this dissertation are are described as fpllows:Firstly, in data preprocessing phase, a fuzzy discretization algorithm based on clustering centers is proposed. This algorithm can be realized in two levels: firstly, using clustering algorithm on the data which needs to be discretized; then confirm the parameters of membership functions, which are used to discretize the data, based on the clustering centers. Through considering the distribution information and class information of data, the algorithm automates the step effectively and overcomes the errors of mining results, which are usually caused by the wrong parameters of membership functions confirmed by man-made.Secondly, in modeling phase, based on the existing algorithm of mining positive and negative association rules, a multiple minimum support method is used, which can ensure the quantity of rules and efficiency of algorithm. Furthermore, minimum correlation coefficient is used to improve the quality of rules. Simulations based on standard data have been done and results showd great effectiveness.Lastly, the marketing application of Hangzhou Telecom PHS package business is introduced. We expatiate on the concept of coss-selling. Then builds the coss-selling model based on association rules mining. Through analysis of historical business data of PHS consumers, the relationship between all kinds of business of PHS is found out and analysed.