Recommend Mobile Phone For Mobile User Based On Data Mining Technology

Mobile communication has become an indispensable part of modern business communication.Mobile communication,satellite communication and fiber optic communication are three new ways of communication.Currently,the development of mobile communication from analog to digital technology stage is moving towards a higher stage of the personal communication.After China's three major 3G operators restructure,China Mobile's advantage is disappearing.In order to retain old customers and attract new users,China Mobile needs to work harder for the sake of customers and provides customers with high-quality service.Based on this,we begin to analysis and mine deep information on all aspects of mobile users.In this paper,we build a recommend system for mobile users through data mining technology.The work of thesis is as follows:1)Clustering main phone models.It includes K-modes algorithm considering initialization based on frequent itemsets(FIK-modes)and clustering algorithm based on a hybrid data with weight.FIK-modes algorithm combines frequent itemsets mining algorithm and K-modes algorithm.Clustering algorithm based on a hybrid data with weight is similar with K-prototype algorithm.It has been normalized and better reflects the significant impact of relationship between attributes of primary and secondary on the clustering results.2)To study the rules which mobile users change mobile phones.It includes the life cycle model and the network of the change of mobile phones.The life cycle model is mainly based on data fitting out a predicted curve.We can find the critical points between maturity and ebb of the predicted curve.The network of the change of mobile phones mining includes key mobile phones mining and leagues of mobile phones.We propose an improved PageRank algorithm to mine the key brands of mobile phones.Leagues of mobile phones divide the network of the change of mobile phones.Which can be used to assist to recommend mobile phones to users.3)Collaborative filtering recommendation algorithm based on users' mixed information.This part includes the algorithm and its cold start problem.Collaborative filtering recommendation algorithm based on users' mixed information firstly finds the nearest neighbors of target users.According to the neighbors,we then calculate the score of each of brands of mobile phones.We finally choose the brands of mobile phones which gain the Top-N highest scores.The cold start problem has two aspects:cold start problem of new users and cold start problem of new brands of mobile phones.