Comprehensible Knowledge Discovery: Gaining Insight from Data

Large databases are routinely being collected in government, science, business and medicine. A variety of techniques from statistics, signal processing, pattern recognition, machine learning, and neural networks have been proposed to help people understand the data by discovering predictive models of the data. However, research in data mining has not paid attention to the factors that make learned models comprehensible. A comprehensible model would provide insight to an expert. This insight could be communicated to others and could support a variety of decision-making tasks. For example, knowledge acquired through such methods on a medical database might be published in scientific journals and provide advice on how to reduce the likelihood of getting a certain disease. Analysis of a political database might reveal the conditions under which trade negotiations are likely to be successful and provide advice on strategies that are effective. Knowledge acquired from analyzing a financial database might be taught in a management school.