|
With the expansion of 3G services, the increasingly fierce competition in the telecommunications market. How to maximize the retention of users in the network, to learn a new customer is one of the most concern of telecommunication enterprises. Competitors' promotions, the introduction of the the company tariff soft landing measures and policies, and regulations are constantly changing, and affect the customer consumer psychology and consumer behavior, resulting in the loss of customer characteristics changing. Shanghai Mobile, the high-end customers is an important component of the mobile revenue, the loss will give the company market share decline, increased marketing costs, falling profits, a series of problems. , Shanghai Mobile's high-end customers is a high-quality customers, the loss of a small user base, is more difficult to capture characteristics. Currently, the average loss of only 1.17% in the high-end customers. In the high-end customer retention rate of Shanghai Mobile is a very important job. According to Plato 2.8 law \The principle of the supremacy can only win. Shanghai Mobile experience to judge the ARPU decline reaches a certain threshold, the number of monthly users call transfer outside call retain. Although this method is able to seize some users, but the whole, less really found the right proportion of users. Early warning model can greatly improve the accuracy of capture the loss of customers, which in the case of limited resources, maximize good customer retention. In this paper, the actual project, the data mining technology in the telecom customer segmentation. Papers with the introduction of the information entropy gain customer segmentation method based on customer behavior, decision tree algorithm to achieve the telecommunications customer segmentation. First, data preparation and data preprocessing and thus differences between several groups of larger and smaller group differences within the group, the result of customer segmentation analysis and target data into customer behavior, future market to evaluate, so as to provide accurate and reliable decision-making guidance. The final design of a customer segmentation analysis systems, general description, cluster analysis, and user management functions, user-friendly customer segmentation, to reduce the loss of high-end users, to ensure the sustainable development of the company's revenue. The algorithm is applied in the actual project, the retention of high-end users in 2010 from an average of 92.3% up to 96.9%, an increase of 4.6%. Estimates, according to the 200 million users of high-end customers to restore to 9.2 million people, to restore the rate to reach 42%. Actual marketing activities, the user using algorithms marketing success rate is nine times the user of the algorithm and prove the effectiveness of the algorithm.
|