The Design And Implementation Of The Recommend System Based On Customer Relations Mining And Score Pretreatment

The mobile device has gradually become one of the main platforms for information with the rapid development of mobile communication technology. However, people face serious overload problems of mobile information with a rich and colorful mobile services and information content increasingly emerging, which is led by the ascension of the mobile multimedia technology and the transmission ability of mobile information.There are two ways to alleviate the information congestion problems effectively in the traditional Internet, which are search and recommendation. Search requires the user take the initiative to conduct information retrieval, so the results depend on the user’s expression ability greatly. However, recommendation is the way, which analysis the information of users and their relationship users to predict their preferences and push results to them, so it doesn’t depend on the users’ability. The resources of mobile communication network are much less than the traditional Internet. What’s more, the ability of the battery and interact is weak for the mobile device. So the method of recommendation is more suitable for searching the information, which users really interested in and enhancing the personalized service experience in the vast information sea.To solve the personalized information content and services recommended issues in mobile communication network, a recommendation algorithm of mixed data selection based on users’social relationship mining and score pretreatment in mobile communication network is proposed with the social network analysis method. It effectively mitigates data sparsely for preferred forecasting and recommendations by using the information of social networks obtained from predicting the potential relationship between social networks and the most similar set of users according to the degree of similar relationship between the users. What’s more, it can pretreat the score matrix before the calculation of the users’similarity to reduce the sparseness of the matrix. The algorithm is proved more accurate and feasible in the experiments by using the public data sets and the simulated data sets.