Finding the best statistical model to predict customer defection in telecommunication retail setting
In this study we examine the question of which statistical mod- els work well in predicting customer defection in the retail mobile telecommunication industry. For each of the two data sets that were used (mobile call pattern and billing, and time taken to churn data), four statistical models were tted and compared namely; arti cial neural networks, decision trees, logistic regression and support vector machines. The arti cial neural network model proved to be supe- rior than the other three models when tted on both data sets. This model gave the best area under the receiver operating characteristic curve (0.93 for call pattern data and 0.88 for billing and time taken to churn data), highest lift at 10 per cent of the population (7.01 for call pattern data and 2.12 for billing and time taken to churn data) and lowest misclassi cation rate (0.04 for call pattern data and 0.19 for billing and time taken to churn data). The logistic regression model under performed the other models when tted to call pattern data and came out as third when tted to billing and time taken to churn data whereby they outperformed the decision tree model. Support vector machine came out as the second best model for billing and time taken to churn data and third when tted to call pattern data. Decision tree model performed well when tted to call pattern data and worst when tted to billing and time taken to churn data The study showed that in the retail mobile telecommunication industry, companies can increase revenue streams and competitive advantage by using data mining techniques to predict customers that are likely to churn. The next step for the business is to embark on retention programs to use these methods to reduce churners.
A research report submitted to the Faculty of Science, University of the Witwatersrand, Johannesburg, in fulfilment of the requirements for the degree of Master of Science in Mathematical statistics. Johannesburg, February 2014.