Applying Models With Scoring
Free (open access)
S Bayerl, T Bollinger & Chr Schommer
\“Scoring”, in general, is defined as the usage of mining models - based on historical data - for classification or segmentation of new items. For example: if the historical data consist of classified customers, then we can use the model for the prediction of the behaviour of a new customer. Scoring offers novel ways to exploit the power of data mining models in everyday business activities, and proliferate mining applications to users who are not educated in mining. In this paper, we present a) the generic scoring process b) its technical implementation, and c) an example of how scoring can be integrated in a real application. The generic process consists of three steps: The mining models are learned first, then they are transferred into the application database, and finally the models are applied to the data loaded in that database. Arguments for the necessity of such a mining improvement are collected. IBM DB2 Intelligent Miner Scoring (IM Scoring) is the first technical implementation of scoring. It is based on the emerging open-standard for mining models (Predictive Model Markup Language - PMML), and the mining extensions for SQL. Implementation issues are discussed, as well as problems that come along with its integration into operational applications. The article closes with the description of a sample application, the integration of scoring into a call center environment. A discussion of the scoring method concludes this article.