首页 理论教育基于持卡人线下消费行为的推荐

基于持卡人线下消费行为的推荐

【摘要】:为此提出了一种基于持卡人线下消费行为的推荐方法。关键词:推荐算法,银行卡,线下消费行为,周期性作者简介:郑建宾,男,1983年生,研究生/硕士,工程师,主管,主要从事及研究领域:数据挖掘、人工智能、电子支付风险防控。

郑建宾

(中国银联电子支付研究院 上海 201201)

摘要:随着电子商务互联网技术的高速发展,推荐系统在电子商务领域得到了广泛的应用,并取得了不错的效果,但在银行卡线下消费场景中并未得到很好的应用。主要因为目前线上的主流推荐算法并不能很好地适应银行卡线下消费场景的特点。为此提出了一种基于持卡人线下消费行为的推荐方法。该方法利用持卡人线下消费周期性的特点,在传统Markov链算法的基础上,认为用户在商户i消费后下一次在商户j消费的转移概率Pij是由上一次用户在和商户i同类型的商户消费到这次在商户i消费的所有记录共同决定的,从而实现持卡人线下消费的商户推荐。实验结果证明,该方法推荐效果良好。

关键词:推荐算法,银行卡,线下消费行为,周期性

作者简介:郑建宾,男,1983年生,研究生/硕士,工程师,主管,主要从事及研究领域:数据挖掘、人工智能、电子支付风险防控。E-mail:zhengjianbin@unionpay.com。

A Recommendation Method Based on Cardholder’s Offline Consumer Behavior(www.chuimin.cn)

ZHENG Jianbin

(Research Institute of Electronic Payment, China UnionPay Co., Ltd., Shanghai 201201, China)

Abstract: With the rapid development of e-commerce and Internet technology, recommender systems have been widely used in the field of e-commerce, and have achieved good results. But in the offline consumption scene, recommender systems have not been well applied. A main reason is that the current online mainstream recommendation algorithm can't well adapt to the characteristics of offline consumer behavior. For this reason, this paper proposes a recommendation method based on cardholder's offline consumer behavior. This method uses the periodic characteristics of the cardholder's offline consumer behavior, realizes the cardholder spending recommended based on the traditional Markov chain algorithm. Experimental results show that the algorithm can provide good recommendation results.

Keywords: Recommendation Algorithm, Bank Card, Offline Consumer Behavior, Periodicity