IMPACT OF HIGH-LEVEL KNOWLEDGE ON ECONOMIC WELFARE THROUGH INTERACTIVE DATA MINING.

This paper describes a novel algorithm for finding the most important relations with the use of data mining. As an example application, the impact of high-level knowledge on economic welfare was analyzed. Our approach, based on interactive data mining, not only helps to discover the most relevant models, but also enables an evaluation of their relevance. The approach is specialized for the analysis of macroeconomic data that often contains incomplete and noisy attributes and, initially, complex relations, where several relations are statistically seemingly important, but only a few are indeed the most relevant. Although data mining algorithms are designed to detect the relevant attributes, irrelevant attributes often appear in the models due to chance-choice, reducing their quality and understandability, and consequently even leading to incorrect conclusions. We present our method and show its application at finding the most relevant relations between high-level knowledge and the state of development of a country. [ABSTRACT FROM AUTHOR]