Customer behavior prediction - it's all in the timing

This paper describes the prediction of customer behavior, which enables service providers in the telecommunications industries to proactively build customer loyalty and to maximize profitability. The prediction process involves extracting customer data, preprocessing the raw data, building a predictive model, and evaluating the model. The modeler needs to understand the service provider's business rules and the application scenario. Customer data used for prediction can be massive and yet sparse, and is dynamic, noisy, and sometimes faulty. For the success of a project, data preparation and timing are often critical components.