Study on a New Method of Rare Pattern Mining Based on QAR Data

Most pattern mining methods concentrate on finding frequent pattern. However, rare pattern mining in some cases is more interesting than frequent pattern mining since rare pattern represents unexpected or unknown information. In this paper, we propose a new method of rare pattern mining based on QAR data to mine rare patterns which is related to pilot unsafe manipulation. The proposed method takes advantage of the correlation between QAR parameter and pilot unsafe manipulation to define the item order in item set and makes use of the concept of global support degree. At the same time, 4 types of focused overrun events are used as entry points to explore pilot unsafe manipulation scenarios in the experiments. The experimental results show that the proposed method in this paper reduces execution time. Finally, pilots and regulatory experts are invited to confirm the mining results. The feedback verify method's feasible and effective.