[c77] S. Daskalaki, I. Kopanas. N. Avouris, Machine Learning Techniques for prediction of rare events in a business environment, Proceedings 3rd Hellenic Conference on Artificial Intelligence, SETN 2004, pp. 79-88, Samos, May 2004. (pdf)
Prediction of rare events in a business environment by using ma-chine-learning techniques presents several difficulties. In this paper, we discuss the design of an intelligent system that predicts rare events. Key aspects of the design process are the class distribution in the training data, use of combined classification algorithms, study of the economic impact of the prediction for the interested company, and combination of predictions with heuristics and data model visualization in the system. The described intelligent system has been developed for prediction of insolvent customers that cannot pay their debts in a telecommunications company. The proposed design process may also be ap-plied in many similar problems, which involve prediction of rare events in other business environments.