Applied Research Of FCM Clustering Based On Genetic Algorithm In The Bank Customer Segmentation

Customer segmentation is a kind of customer classification processes, which divides the customers of a certain product into a number of different parts based on customers'information collected and collated by enterprises according to various significant differences in the customers'(including consumers and distributors) demand characteristics, buying behavior, buying habit, reputations and so on. Clustering analysis is one of the primary means of customer segmentation as well as a multivariate analysis using quantitative mathematics to determine the affinities of samples so as to make an objective classification. However, the traditional clustering analysis is a kind of mechanical division, which strictly divides each of the individual objects to be identified into a class with the nature of critical definition, so the boundary of this category is defined clearly. In fact, the majority of clients don't have a strict property and the classification of their behavior is somewhat intermediate, which suits with soft breakdown. Fuzzy set theory provides the soft breakdown with a powerful analytical tool. Therefore, this paper introduces the fuzzy C-means clustering algorithm into the progress of customer segmentation, and uses MATLAB programming algorithm to achieve the calculation.The results of the fuzzy C-means clustering algorithm are readily to fall into local minimum due to being affected by the initial clustering center, while the genetic algorithm has the ability of global optimization. Therefore, this paper combines the genetic algorithm with the FCM clustering algorithm by using the genetic algorithm to initialize the FCM clustering center, which reduces the sensitivity of the FCM clustering algorithm to the initial value and also makes the FCM clustering algorithm achieve global optimization. In addition, the number C of clustering for the FCM clustering algorithm must be given in advance, however, in front of the large numbers of data, it is often impossible to distinguish the discrete data, not to mention the division, so a given number of clustering may lead to a wrong category, and make the clustering unreasonable. Therefore, the paper uses the effective indicators of fuzzy clustering to find the best clustering C so as to use fuzzy C-means algorithm and genetic-FCM clustering algorithm to segment the customers of a certain bank based on the bank's customer information, and compares these two algorithms as well.。...