SUGI 27: To Neural or Not to Neural? -- This Is the Question

Adding new lines of business and new services for existing customers is particularly important in deregulated telecommunication industry. Attracting new services to existing customers often translates into happier customers, increased retention of profitable customers, and competitive advantage. Identification of the most profitable customers (who are the most probable buyers of additional or new services) calls for computer based systems that are based on data mining solutions. This paper compares solutions from neural networks to solutions from logistic regression, discriminant analysis, and decision tree analysis for cross selling predictions. This comparison deals with the statistical accuracy of predictions, ability for business interpretation, and the practical implementation of computer systems that can perform outlined tasks. Included in the paper are practical tips on how to use base SAS, SAS/STAT, and SAS Enterprise Miner for model selection and the decision support systems (DSS) implementation.