Some Recent KDD-Applications at DaimlerChrysler AG

The mission of the Information Mining department, DaimlerChrysler Research and Technology, is exploring, exploiting, and enriching data mining and text mining methods to provide complex decision and product support systems. A key requirement for being able to provide a complex system lies in the different core technologies used, such as symbolic machine learning, statistical learning procedures, association learning, neural networks, distributed data mining, text mining, and model selection procedures. On the basis of these core technologies, we extract and analyze information from data collected in vehicles, from business data, financial data, and documents. Awareness of the above-mentioned technologies together with know-how about other topics like optimization and case-based reasoning build only one part of the expertise of our research department. The complementary part consists of knowledge about different application domains such as computational marketing, computational finance, car market modeling, data cleaning, and knowledge from various technical domains. In this paper some topics of the second part are presented. 1 Prediction of warranty and goodwill costs 1.1 Problem Description The warranty planning in the automobile industry is characterized by several problems: the need to calculate planning figures for cash reserves within the balance caused by legal or voluntary obligations in the context of warranty and goodwill a growing number of actually manufactured vehicles and a growing number of production plants all over the world, resulting into a more complex and multilayered production structure a different warranty and goodwill policy for different sales markets. The system WAPS, that is the result of a joint project between DaimlerChrysler Research & Technology and the Sales and Services Department, should master the business needs