Prediction Technology Of Lost Customers In The It Industry Based On Decision Tree Algorithm

With the foreign IT companies seek to enter the Chinese market, European companies have the rich advantages of both product technology and marketing management experience, So that the already saturated domestic IT market more competitive. Domestic IT companies have to face competition from European and American IT companies to increase the new market share in increasingly difficult and the loss of old customers caused great loss of falling profits, while the overall costs of acquiring new customers far higher than the cost of maintaining old customers, To some extent have affected the healthy development of domestic IT companies.How to find potential customers prior to the loss of the possibility of the losing customers, to target specific customer groups to take the appropriate marketing strategy, retain old customers, which have become a common problems faced by IT companies. Data mining is to extract out the implicit knowledge and information from the vast amounts of data which people's attention, to some extent, data mining solves the problem of IT industry customer churn.This paper completed the major works are as follows:(1) Analysis and research the data mining classical decision tree algorithm, described the decision tree principle and process of formation, analysis the advantages and disadvantages of the commonly used decision tree algorithm ID3,C4.5 and C5.0 algorithm.(2) Using Clementine data mining tools, combined with C5.0 decision tree algorithm to build customer churn prediction model, based on customer churn prediction model established to evaluate and analyze.(3) Complete the loss of customers in the system framework and the system features modular design based on functional modules to achieve customer churn prediction. Finally, Shi Heng technology customer data as an example database, comparative and analysis of historical customer churn prediction system over the same period before and after using the actual amount of customer churn, the actual number of customers than the actual loss of history reduced to some extent over the same period.This paper combining the data mining theory and project practice, and ultimately the prediction system will be applied to the possibility of a potential loss of customer identification. The results show that the established customer churn prediction model is basically meet the demand of IT industry customer churn prediction, to some extent, which solved the problem of higher loss rates of the IT enterprise and achieved the expected goals. Customer churn prediction system has a theoretical meaning and practical value.