Research Of Clonal Selection Algorithm Focused On Customer Relationship Mining

Customer Relationship Management is a hot issue of the enterprise management currently. Especially, the research and application of CRM model for demand forecast based on data mining has a positive role in promoting growth on business profits. Therefore, it attracts ever-increasing great attention of companies. In order to forecast potential customers, the intelligent customer relationship management system often requires comprehensive analysis. Most existing studies on potential customers utilize traditional classification methods. But the traditional classification methods often have some defects, such as low efficiency, easily affected by system parameters, poor interpretability and so on. To solve these problems, referred to the immune memory mechanism, this paper takes the advantage of the clonal selection algorithm to built a classifier to predict potential customers.This paper applies the classical clonal selection algorithm as the computing framework. It studies all aspects of clonal selection algorithm and incorporates the ideas of clonal selection, clonal mutation, antibodies complement and so on. It improves the clonal selection algorithm by incorporating the affinity of antibody into mutation part. The revised algorithm adjusts the mutation step according to the affinity of antibody after each iteration. The cloned antibodies have the ability to keep the characteristics of the parent antibody as much as possible. Furthermore, it improves the convergence rate by searching around the antibodies with higher affinity, so that there is higher probability to find the optimal antibody. Besides, to prevent antibody degradation, it takes the advantage of clonal deletion to replace the antibodies with lower affinity in the memory antibody population with the candidate set. Additionally, the algorithm improves the diversity of antibody by utilizing the antibody mutation. Therefore, the revised algorithm achieves the optimization of search in the global scope, and avoids falling into local search. By these operations, it enhances the performance of optimization of searching in a partial or global scale.On the other hand, based on immune memory mechanism, this paper utilizes the memory antibody population after clonal selection as the classifier to classify the data set. After a series of processing phases, such as cloning, mutation, selection and supplement, this paper obtains memory cell groups from the training set and classifies the test set with the KNN classification method. As the experiments with UCI data sets indicate, the classifier demonstrates a desirable classification results.Finally, based on the actual demand of an enterprise CRM and incorporated classification methods in data mining, this paper designs the classifier with an improved clonal selection algorithm and implements a prediction system for potential customers. This system solves the CRM in the excavation, identification forecasts and etc. And the performance demonstrates a desirable result.