Identify error-sensitive patterns by decision tree
- Publication Type:
- Conference Proceeding
- Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2015, 9165 pp. 81 - 93
- Issue Date:
© Springer International Publishing Switzerland 2015. When errors are inevitable during data classification, finding a particular part of the classification model which may be more susceptible to error than others, when compared to finding an Achilles’ heel of the model in a casual way, may help uncover specific error-sensitive value patterns and lead to additional error reduction measures. As an initial phase of the investigation, this study narrows the scope of problem by focusing on decision trees as a pilot model, develops a simple and effective tagging method to digitize individual nodes of a binary decision tree for node-level analysis, to link and track classification statistics for each node in a transparent way, to facilitate the identification and examination of the potentially “weakest” nodes and error-sensitive value patterns in decision trees, to assist cause analysis and enhancement development. This digitization method is not an attempt to re-develop or transform the existing decision tree model, but rather, a pragmatic node ID formulation that crafts numeric values to reflect the tree structure and decision making paths, to expand post-classification analysis to detailed node-level. Initial experiments have shown successful results in locating potentially high-risk attribute and value patterns; this is an encouraging sign to believe this study worth further exploration.
Please use this identifier to cite or link to this item: