Study Of CRM System Based On Decision Tree Classification Algorithms

In this paper the analysis and comparison on data mining and prediction algorithm were made base on which the stability and efficiency of Decision Tree were analyzed. A kind of optimized algorithm named Q-ID3 was put forward, which aimed at the flaws of value incline in traditional ID3.The formula of the prior knowledge classification standard was deduced used of the information entropy and the condition entropy formula, which add the priori knowledge parameter to the weighting sum. This parameter Q ranges from 0 to 1 and its value is determined by use of prior knowledge and realm knowledge. The parameter plays a role in strengthening the weight of important attributes and decreasing the weight of unimportant attributes. For a large instance set, Q-ID3 algorithm can reduce search space and improve learning efficiency, build simple structure decision tree, which is accordant with actual circumstance on the basis of prior knowledge parameter. Q- ID3 algorithm reserved simple classification method and intelligible rule of the traditional ID3 and avoided that attributes which have large search space and are not always the best are adopt as test attributes.Comparing the traditional ID3 algorithm and the modified ID3 algorithm, we find although our method, which generates dataset randomly with various sizes, is time consuming in the initial stage of the process in order to obtain the parameters, as the size of data set increases, the rapid growth of the decision tree makes the whole process less time consuming while the structure of the generated decision tree based on our approach is more precise and efficient than the traditional approach. This shows that our method outperforms the traditional approach in a considerable way.The data mining was made on the customer data of China Mobile by use of Q-ID3 algorithm. A decision tree was constructed according to the prior knowledge form communication market to analyze the customer loss, which provide decision base for CRM to reserve consumers and prevent consumer loss.