The Research on Broadcast Television User Dividing Groups Technology Based on Concept Data Clustering Ensemble

Aiming to meet the demand of intellectual delivery business, this paper puts forward Broadcast Television user dividing groups technology based on concept data clustering ensemble. Firstly, by Glass data, Blance data and Zoo data in UCI, the paper verify that concept data clustering ensemble technique based on K-MODES method can get more stable and more accurate results than K-MEANS method. Secondly, we calculate the audiences´ viewing preferences according to the ratings data, and get the multiple classification results by K-MEANS Clustering method. We take the clustering group as a concept data set, transform the problem of consensus function in clustering ensemble to a common clustering problem, and apply the concept data clustering algorithm to get a unified clustering result, so as to achieve Broadcast Television user dividing groups.