We present a novel approach for using the pattern position distribution as features to detect software failure. In this approach, we divide an execution sequence into several sections and compute the pattern distribution in each section. The distribution of all patterns is then used as features to train a classifier. This approach outperforms conventional frequency based methods by more effectively identifying software failures occurring through misused software patterns. Comparative experiments show the effectiveness of our approach.
|Journal||International Journal of Computational Intelligence Systems|
|Publication status||Published - 1 Mar 2013|
- Sequential Patterns
- Classification Algorithm
- Software Failure
- Anomaly Detection