Choosing data-mining methods for multiple classification: representational and performance measurement implications for decision support

摘要: Data-mining techniques are designed for classification problems in which each observation is a member of one and only category. We formulate ten data representations that could be used to extend those methods observations may full members multiple categories. propose an audit matrix methodology evaluating the performance three popular data-mining techniques--linear discriminant analysis, neural networks, decision tree induction-- using technique can accommodate. then empirically test our approach on actual surgical set. Tree induction gives lowest rate false positive predictions, version analysis yields negatives category problems, but networks give best overall results largest cases. There substantial room improvement all techniques.