An Efficient Algorithm for Mining High Utility Contiguous Patterns from Software Executing Traces

Software behavior pattern mining has important significance since it can provide help for software engineers to maintain the correctness of software and detect exceptions as soon as possible. These high utility software behavior patterns shed light on software behavior and capture unique characteristic of software traces. In this paper, we propose a novel approach HUCP-Miner (high utility contiguous pattern mining) to mine high utility contiguous patterns from the software executing traces. First of all, this work presents a maximum utility measure which is used to simplify the utility calculation for contiguous patterns. Second, we propose a novel structure called UL-list (utility and location list) to store utility and location information of patterns which contributes to backward extension. Based on UL-list, a remaining utility upper bound model (ruub) and extension strategy are put forward to prune the unpromising patterns early. Finally, an extensive experimental study with different real-life datasets shows that the proposed algorithm has impressive performance.