Investigating human factors issues in the data mining processes

In our information age, a rapidly widening gap between data collection/storage capabilities and the ability to analyze the data has eliminated the feasibility of traditional data analysis approaches. The root of the problem has been identified as data size and its dimensionality (Fayyad, Haussler, Stolorz, 1996). Data Mining (DM) has become part of the technological response to this issue. The promise of DM into find any interesting patterns hidden in these gigabytes of data and present the extracted high-level information to the analyst or decision-maker. Despite the fact that human involvement in DM tasks tends to be significant (e.g., Romanoski, 1999; Ankest et al., 2000), there has been a prejudice against the need for, or importance of, the human in the DM process because of over emphasis on its "automatic" or "semiautomatic" aspects.; To the best of our knowledge, the current study is the first one to systematically investigate human roles in the DM process and model the DM system (in fact, a human computer system) as a dynamic and interactive whole with four individually functioning components. In addition, as the first attempt from outside the DM community to study human issues in the DM process, the current study has introduced useful methodology that is specifically designed to study the human computer system.; The current study employed a progressive modeling approach with data from published studies, interviews, and field observations with Subject Matter Experts (SMEs). Two sets of ACTA questionnaires invited eight SMEs to introspect the mental demands they faced and cognitive skills they applied in their DM activities. The first part of field observation studied six SMEs applying DM techniques in their research, identified and analyzed their interfaces with the computer during the DM process, decision making points, and complexity and resources in the decision making. Their struggles and concerns about the current DM system and how they have been accommodating to the cognitive demands were elicited and documented. The second part of field observation focused on collaborative meetings between domain experts and data mining experts. These meetings became zoomed views of those sub tasks found in the stage of selecting or developing algorithms to build a DM model. The collaborative meetings revealed the socio-technical aspects of the DM system, and directed further investigation on how human data analysts with different expertise communicate and integrate their knowledge and skills in developing DM models. (Abstract shortened by UMI.)...