Research On Fusion Of Ant Colony Clustering Algorithm And K-means Clustering Algorithm And Its Application On Customer Segmentation

Now the market competition is becoming increasingly fierce. Due to the huge number of consumers, the public traditional marketing methods for companies are not only expensive but also in disadvantage. Companies can do accurate marketing for different user groups to improve the efficiency of marketing activities and get better marketing results by using customer segmentation method. Faced with the large amount of user data that cannot be analyzed by the tradition al data analysis methods, companies need the data mining techniques. By utilizing such method that is more suitable for large amounts of data analysis and processing, companies can achieve customer group segmentation.In this thesis, we explained the theory of marketing principles and fundamental of data mining techniques. We also researched the basic principles of K-means clustering algorithm and ant colony clustering algorithm. By analyzing the user data from the bbs of Changhong, according to the character of large amount and variance in attributes of use data, this thesis proposed an improved algorithm that is suitable for forum user clustering, which is named Ant-K-means clustering algorithm. Furthermore, this thesis not only discussed the way of choosi ng cluster quantity and the quality sensitivity of initial cluster center for K-means clustering algorithm, but also considered the solution for ant clustering algorithm’s long searching time problem. By utilizing such algorithm on sample data clusters fro m central data, and then by employing the cluster amount and center as input parameters, and later by using the K-means Algorithm on all data, based on pheromone we modified in this thesis, the final conclusion for clustering is drawn. By applying the modi fied clustering algorithm on forum of Changhong intelligence TV, according to various customers’ behavior data, we can categorize them into groups for analysis and visualization.Applying the modified clustering algorithm on the different groups of customers from bbs of Changhong, the results appeared excellent proved by the high similarity of customers’ characters in same segmentation. The result verified the correctness and effectiveness of the algorithm, and provided basic fundaments for Changhong intelligence television’s marketing and customer services.