Towards the Personalization of Algorithms Evaluation in Data Mining

Like model selectionin statistics,the choiceof appropriate Data Mining Algorithms (DM-Algorithms) is a very importanttask in the processof KnowledgeDiscovery.Due to this fact it is necessaryto havesophisticatedmetricsthat can be used as comparatorsto evaluatealternativeDMalgorithms. It has been shown in literature, that Data EnvelopmentAnalysis(DEA) is an appropriateplatformto developmulti-criteria evaluationmetricsthat canconsiderin contrary to mono-criteria metrics - all positive and negativepropertiesof DM-algorithms. We discuss different extensions of DEA that enable considerationof qualitative properties of DM-algorithms and considerationof userspreferencesin developmentof evaluationmetrics.The resultsopennew discussionsin the generaldebateon modelselectionin statisticsandmachine learning.

• Type: Conference
• Year: 1998
• Where: KDD
• Authors: Gholamreza Nakhaeizadeh, Alexander Schnabl
• Keywords: Appropriate Data Mining, Data Mining, KDD 1998, Model Selectionin Statistics, The