Journal of Computer Applications ›› 2016, Vol. 36 ›› Issue (7): 1904-1908.DOI: 10.11772/j.issn.1001-9081.2016.07.1904

Previous Articles     Next Articles

Customer behavior prediction for card consumption based on two-step clustering and hidden Markov chain

SONG Tao1, WANG Xing2   

  1. 1. School of Statistics, Capital University of Economics and Business, Beijing 100097, China;
    2. Applied Statistical Research Center & School of Statistics, Renmin University of China, Beijing 100872, China
  • Received:2016-03-09 Revised:2016-05-03 Online:2016-07-10 Published:2016-07-14
  • Supported by:
    This work is partially supported by the Fundamental Research Funds for the Central Universities (13XNI011).


宋涛1, 王星2   

  1. 1. 首都经济贸易大学 统计学院, 北京 100097;
    2. 中国人民大学 应用统计研究中心&统计学院, 北京 100872
  • 通讯作者: 王星
  • 作者简介:宋涛(1971-),男,北京人,博士研究生,主要研究方向:数据挖掘、经济统计分析;王星(1971-),女,北京人,副教授,博士,主要研究方向:数据挖掘、统计学分析。
  • 基金资助:

Abstract: Bank card payments account for a large proportion in the social consumption, which plays a major role in the promotion of economic growth. So, predicting consumer behavior is important. However, the traditional methods are difficult to effectively deal with complex data and dynamic changes. Based on this, a customer behavior prediction method for card consumption based on two-step clustering and Hidden Markov Chain (HMC) was presented. Firstly, consumer behaviors were conduced by pattern clustering based on sequence; then the secondary clustering was conducted by introducing penalty clustering, which carried out the equilibrium division of the hierarchical states in the sequential pattern. Secondly, HMC was used to estimate the state transition of consumption levels in the sequence and predict the future consumer behavior of the users. Finally, the experimental comparison and analysis results on the traditional clustering, clustering without penalty and clustering with penalty show that the proposed method based on two-step clustering and HMC is more suitable to the consumer behavior prediction model.

Key words: two-step clustering, penalty clustering, Hidden Markov Chain (HMC), card consumption, behavior prediction

摘要: 银行卡支付在社会消费行为中占很大比例,在促进经济增长中发挥重大作用,因此,预测持卡消费行为具有重要意义。然而,传统方法难以有效应对复杂数据和动态变化。为此,提出基于二次聚类和隐马尔可夫链(HMC)理论的个体消费行为预测方法。首先,对消费行为按照序列进行模式聚类,并引入惩罚聚类进行二次聚类,对序列模式中的层次状态进行平衡划分;其次,利用HMC来估计序列中消费层次的状态转移,对用户的未来消费行为进行预测。最后,通过实验比较分析传统聚类、无惩罚序列聚类和带惩罚项的聚类结果表明,提出的基于二次聚类和隐马氏链的方法更适用于消费者行为预测。

关键词: 二次聚类, 惩罚聚类, 隐马尔可夫链, 持卡消费, 行为预测

CLC Number: