Customer Churn Analysis In Mold Industry Based On Data Mining

Mold is a product which was manufactured according to the order that including customer needs, brutal competition lead to the profit of mold products is declining, it is critical for mold enterprise to maintain the bottom line of profits that reduce customer churn rate. Mold manufacturing is a creative process based on experience and knowledge, and the quality of service provided by mold enterprise in this process is the core competitiveness to win customers. So analysis the reason for churn and improve service quality is not only critical for mold enterprise’s success, but also an important research topic having practical significance. Applying data mining technology to analysis mold customer churn is the theme, the main content of the essay include the following aspects.First of all, clear research ideas for mold customer churn analysis. The paper selected data mining as the technique for customer churn analysis, after analyzing the need for mold customer segmentation, the paper divided the mold customers based on customer present value and potential value according to the characteristics of mold industry, and selected high-value customers as the target for churn analysis. Meanwhile, the essay reviewed the concept of customer churn, in order to evaluate the level of customer churn accurately, a parameter called customer churn degree was defined based on changes in mold customer’s order, it provided theoretical groundwork for the following research.Secondly, analyzed the method of customer churn analysis in mold industry.The essay used statistical correlation technique to filter attribute variables so as to pick up the key attributes which have a greater impact on customer churn as the input of model. Then, after a comprehensive analysis about the applicability of common methods in data mining, the paper selected decision tree C4.5algorithm as a benchmark according to features of actually business data in mold enterprise. In addition, considering the importance of domain knowledge when analyze the reasons for churn in mold industry and the shortage of traditional C4.5algorithm that cannot bring in domain knowledge during modeling, a method that incorporate domain knowledge in decision tree model through applying monotonicity constraints was proposed, so as to pruning the decision tree by domain knowledge obtained from mold experts.At the end, a case study about customer churn analysis in mold industry.The paper selected a big mold company as research subject, the sample data sets composed by customer information and order data were obtained from it, after filtering attribute and data preprocessing, a customer churn prediction model was established by C4.5algorithm, then the decision tree was pruned through domain knowledge obtained, finally, a contracted model was build. By comparing with the model build by traditional decision tree, results showed that the proposed method not only maintained the same high prediction accuracy, but also improved the comprehensibility and rationality of the model in a large degree. In addition, the essay explained the rule sets that extracted from final model.The paper established mold churn prediction model by using data mining according to the characteristics of the mold industry churn analysis, at the same time, a method that incorporate domain knowledge in decision tree model through applying monotonicity constraints was proposed.The conclusion of the paper not only provide effective decision-making information for mold companies to analysis the reason for churn, but also provide a reference on churn analysis in mold industry through data mining technique.