Developing knowledge structures: A comparison of a qualitative-response model and two machine-learning algorithms

摘要

Many unstructured decision-making problems are too complex to be solved by existing analytical methods. Organizations commonly rely on human expertise to solve such problems. Human expertise is presumed to utilize specialized knowledge structures, and recent research in artificial intelligence has sought to devise machine-learning procedures that can derive knowledge structures from sets of data that represent problem-domain exemplars. Concurrently, research in statistical methods has resulted in the development of qualitative-response models that can be applied to some of the same kinds of problems addressed by machine-learning models-especially those that involve a classification decision. Using a difficult audit decision problem requiring expertise, this paper reports on a study comparing performance of two machine-learning algorithms and logistic regression, a widely-used qualitative-response method for classification analysis. A recently-developed machine-learning algorithm, NEWQ, is found to yield the best performance characteristics.