Attribute Selection for Telecommunication Churn Prediction

The telecommunication industries customer’s bases are increasing every day. The industries are expected to significant loss of income due to increasing competition in drawing customers towards their customer bases. It is important to find the cause for losing customers and identifying the importance of the customer and retain them. The customer leaving the present telecom customer base and moving to other telecom service providers is called as churn. The telecommunication data set considered for identifying the importance of customer and churn prediction contain high dimensional data, it may contain redundant and inappropriate attributes. To apply the data mining tasks it is difficult to deal with high dimensional data and it leads to inappropriate predictions. To apply data mining task it is necessary to pre-process the data and reduce high dimensional data to low dimensional data without losing the prediction information. The reduced low dimensional data gives best results in churn prediction. This work focus on different attribute important measures and selection methods for identify the best subset of attributes for churn prediction. The experimental results of different attribute selection methods produces significant subset of attributes from high dimensional telecom dataset. The approach proven that it is helpful for predictive accuracy of further telecom churn management.