基于先验频繁模式算法的航班延误关联规则挖掘

Flight delay association rules mining based on Apriori algorithm

为探究航空公司航班延误特性,利用基于约束-聚类的多维数据预处理方法,采用先验频繁模式算法挖掘运行要素和延误事件之间的内在关联。选取某航空公司连续3年的北京—广州航班运行数据,通过频繁模式搜索和关联规则挖掘,经置信度和支持度阈值的筛选,选取提升度大于1的强关联规则形成规则库;分析研究了包含前项数为1和2在内的32条强关联规则,并使用真实数据进行了饱和性和有效性验证。研究结果表明:航班延误关联规则库的判断准确性较高,达到了86.7%;极端气象条件、前序航班延误状态、航班计划时刻和其他部分时间属性,显著增大了航班延误现象出现的概率;此外,对特定属性要素组合发生的潜在作用的挖掘,证实了对航班运行控制决策参考的有效性。

In order to explore the flight delay characteristics of airlines,the multi-dimensional data preprocessing method based on constraint clustering and Apriori algorithm were used to mine the internal association between operation elements and delay events in this paper.The data of flight operations from Beijing to Guangzhou of an airline for three consecutive years were selected to mine the association rule.After the screening of confidence and support threshold,the strong association rules with lift greater than 1 were selected,and 32 strong association rules including the number of previous items of 1 and 2 were extracted for research and analysis.The saturation and effectiveness were verified by using the real data.The results show that the judgment accuracy of flight delay association rules is about 86.7%and extreme weather conditions,previous flight delay status,flight schedule and other time attributes significantly increase the probability of flight delay.In addition,the mining of the potential role of the combination of specific attribute elements confirms the effectiveness of the reference for flight operation control decision-making.