The Application Of Objective Cluster Analysis In Customer Segmentation

Customer segmentation refers that a company can classify it’scustomer in a specific market, according to customers’ attributes,behavior, needs, preferences and value, then offer different products,service and sales model. The value that customers provide to anenterprise varies from person to person, therefore, while managingcustomers, a company need to do some statistics, analysis or somecustomer segmentation to her customers. Only in this way, could acompany do some different marketing activities according to customers’characteristics, maintain and try to expand customers with high value,mine and develop potential customers and give up low value customers. Itis the only way for a company to expand its market and benefit more todo effective customer segmentation.Today, the world’s economic is developing rapidly and retailindustry is becoming more and more open, it is very hard for a companyto win any customer with low cost, more and more enterprises begin topay attention to customer-centric, they begin to attach more importance tocollect data about their customers, such as the basic information ofcustomers, consumer records and other information, then try to dig someuseful message from this. Customer segmentation is one kind ofenterprise’s’ data mining.Cluster analysis is very extensive in the application of customer segmentation, and traditional algorithm like K-means, CURE, STINGand SOM is frequently used clustering algorithm, this methods are easy touse, but there are also defects for this algorithm: the user must intervenebefore modeling. For example, while clustering with K-means, thenumber of category is confirmed by modelers, even though they haveunusually rich experience in marketing, this algorithm is too dependenton them, the number of clusters would inevitably too much or too littleand cannot objectively response the customer features.In this paper, I did a thorough comparative study on the performanceof the traditional clustering algorithms, compared these algorithms fromdifferent angles, and then pointed the advantages and disadvantages oftraditional clustering algorithms. Based on this, this paper used a newclustering method——the objective cluster analysis, this method avoidsthe subjective effects of modelers and the results can response thecharacteristics of customers objectively. We applied it to customersegmentation for retail industry, compared with the traditional K-meansalgorithm, the OCA(Objective cluster analysis) showed its advantages inclustering.