Abstract: Aiming at the characteristics of large amount of fault data and rapid growth in the after-sales service of automobile industry chain platform, and the defects of traditional FP-growth algorithm in processing massive data, an improved FP-growth algorithm based on MapReduce is proposed to mine the association relationship in the after-sales fault information. The algorithm combines the advantages of pruning strategy and balanced grouping strategy. The pruning strategy is used to reduce the number of iterations of item set mining. Based on the balanced grouping algorithm, the load balancing of parallel frequent pattern mining process is realized. The experimental results show that the proposed algorithm performs better. Taking the historical fault data of the after-sales service of automobile industry chain collaborative platform as a sample, the important faults with high frequency of occurrence and the associated faults with high probability of simultaneous failure are discovered.

Key words: automobile after-sales service, fault data, FP-growth improved algorithm, pruning strategy, balanced grouping

摘要： 针对汽车产业链平台售后服务业务中故障数据量大、增长速度快的特点，以及传统FP-growth算法在处理海量数据时的缺陷，提出了基于MapReduce的FP-growth改进算法挖掘汽车售后故障信息间的关联关系。算法同时结合剪枝策略和均衡分组策略的优势，采用剪枝策略减少项集挖掘的迭代次数，基于均衡分组算法实现并行频繁模式挖掘过程的负载均衡。实验结果表明提出的算法性能较优。以汽车产业链协同平台的售后服务历史故障数据为样本，挖掘得到出现频率较高的重要故障件，以及同时发生故障概率较大的关联故障件。

关键词: 汽车售后服务, 故障数据, FP-growth改进算法, 剪枝策略, 均衡分组