深圳大学学报理工版

针对通过零售交易数据进行客户分群时传统方法未考虑商品的价值问题,提出用RFM(recency frequency monetary)表达交易数据的方法,该方法将客户购买的商品和商品类别组成一棵RFM购买树(recency frequency monetary purchase tree,RFMPT).提出基于RFM购买树的快速聚类算法(based recency frequency monetary purchase tree clustering,BRFMPTC),把购买树构建为CoverTree(CT)索引结构,利用CT结构快速选择k个密度最大的购买树作为中心,将其他对象划分到距它最近的类中心. 实验结果表明,在距离加权下,BRFMPTC算法较传统算法在整体上能产生质量更高的聚类结果,性能得到较大提升.

In order to solve the problem that the value of goods has not been considered in traditional methods of customer segmentation, we propose a method of using the recency frequency monetary purchase tree(RFMPT)to represent transaction data, in which a RFM purchase tree is built based on the category of the goods.Based on the RFM purchase tree,we propose a fast clustering algorithm named based recency frequency monetary purchase tree clustering(BRFMPTC). This algorithm constructs the purchase tree as a CoverTree(CT)index structure. With this structure, we can quickly select the k densest purchase trees as cluster centers, then divide the other objects into the nearest class center.The experimental results show that the performance of the proposed method with distance weighting is better than that of the traditional clustering algorithms.

引言
1 RFM购买树表达及相似度计算
2 基于RFM购买树表达的聚类算法
3 结果与分析
4 结语

图1 RFM购买树示意图<br/>Fig.1 A sample of RFM purchase tree

图1 RFM购买树示意图
Fig.1 A sample of RFM purchase tree

图2 CreateRFMPurTree算法<br/>Fig.2 CreateRFMPurTree algorithm

图2 CreateRFMPurTree算法
Fig.2 CreateRFMPurTree algorithm

图3 CoverTree示意图[12]<br/>Fig.3 A sample of CoverTree[12]

图3 CoverTree示意图[12]
Fig.3 A sample of CoverTree[12]

图4 EstimateDensity算法<br/>Fig.4 EstimateDensity algorithm

图4 EstimateDensity算法
Fig.4 EstimateDensity algorithm

图5 BRFMPTC算法<br/>Fig.5 BRFMPTC algorithm

图5 BRFMPTC算法
Fig.5 BRFMPTC algorithm

表1 各数据集大小和平均稀疏度<br/>Table 1 The size of each data set and the average sparsity

表1 各数据集大小和平均稀疏度
Table 1 The size of each data set and the average sparsity

图6 不同k下各算法lg(W(k))值<br/>Fig.6 (Color online)lg(W(k)) of each algorithm under different kDI

图6 不同k下各算法lg(W(k))值
Fig.6 (Color online)lg(W(k)) of each algorithm under different kDI

图7 各算法在不同r和k下的DI值<br/>Fig.7 (Color online)The DI value of each algorithm in different r and k

图7 各算法在不同r和k下的DI值
Fig.7 (Color online)The DI value of each algorithm in different r and k

图8 各算法在数据集D1到D10上的运行时间<br/>Fig.8 (Color online)The execution time of each algorithm

图8 各算法在数据集D1到D10上的运行时间
Fig.8 (Color online)The execution time of each algorithm

[1] Natchiar S U, Baulkani S. Customer relationship management classification using data mining techniques[C]// International Conference on Science Engineering and Management Research. Dubai: IEEE, 2014:e18.
[2] Chattopadhyay M, Dan P K, Mazumdar S, et al. Application of neural network in market segmentation: areview on recent trends[J]. Management Science Letters, 2012, 2(2):425- 438.
[3] Chen Muchen, Chao Chuangmin, Wu Kuanting. Pattern filtering and classification for market basket analysis with profit-based measures[J]. Expert Systems, 2012, 29(2):170-182.
[4] Böttcher M,Spott M,Nauck D, et al. Mining changing customer segments in dynamic markets[J]. Expert Systems with Applications: An International Journal, 2009,36(1):155-164.
[5] Müller H, Hamm U. Stability of market segmentation with cluster analysis:a methodological approach[J]. Food Quality and Preference, 2014, 34(2):70-78.
[6] Singh A, Rumantir G, South A, et al. Clustering experiments on big transaction data for market segmentation[C]// Proceedings of the 2014 International Conference on Big Data Science and Computing. New York, USA: ACM, 2014:1-7.
[7] Zalaghi Z, Varzi Y A. Measuring customer loyalty using an extended RFM and clustering technique[J]. Management Science Letters, 2014, 4(5): 905-912.
[8] 李刚,张莉,李纯青.交易数据的百货商场客户细分研究[J].西安工业大学学报, 2014(3):216-220.
[9] Hsu F M,Lu Lipang,Lin Chunmin. Segmenting customers by transaction data with concept hierarchy[J]. Expert Systems with Applications: an International Journal,2012,39(6):6221- 6228.
[10] 蔡玖琳,张磊,张秋三.一种基于数据挖掘的零售业客户细分方法研究[J].重庆工商大学学报自然科学版, 2015,32(2):43- 48.
[11] Chen Xiaojun,Huang Joshua,Luo Jun.PurTreeClust: a purchase tree clustering algorithm for large-scale customer transaction data[C]// IEEE 32nd International Conference on Data Engineering. Helsinki: IEEE, 2016:661- 672.
[12] Beygelzimer A, Kakade S, Langford J. Cover trees for nearest neighbor[C]// Proceedings of the 23rd Inter-national Conference on Machine Learning.Pittsburgh, USA: ACM, 2010:97-104.
[13] Hall M, Frank E, Holmes G, et al. The WEKA data mining software: an update[J]. ACMSIGKDD Explorations Newsletter, 2009, 11(1):10-18.
[14] Sander J, Ester M, Kriegel H P, et al. Density-based clustering in spatial databases: the algorithm GDBSCAN and its applications[J]. Data Mining and Knowledge Discovery, 1998, 2(2):169-194.

备注

引言

1 RFM购买树表达及相似度计算

2 基于RFM购买树表达的聚类算法

3 结果与分析

4 结语

期刊信息

备注

引言

1 RFM购买树表达及相似度计算

2 基于RFM购买树表达的聚类算法

3 结果与分析

4 结 语

期刊信息

4 结语