Root-Cause Analysis with Interactive Decision Trees
Alexander Detzner, Ramona Ruckschlos, Martin Eigner · 2020 24th International Conference Information Visualisation (IV) · 2020
When manufacturing products fail, root-cause analysis is essential to improve the affected products and processes. However, finding the relevant influences among numerous possible factors can be a challenge. Computer-aided methods for anomaly detection and pattern recognition can significantly speed up this process to provide great benefits in quality management applications. We introduce a method for root-cause analysis employing interactive decision trees. Improving upon an existing concept, we developed a software prototype that utilizes a lean yet powerful representation of relationships and influences. Several interactive options allow quality data analysts to integrate their expert domain knowledge into the pattern recognition process and quickly identify anomalies among the product attributes of defective vehicles.