基于缺失数据BN参数学习的电信流失客户预测算法

doi:10.11959/j.issn.1000-0801.2018018

Abstract

Abstract:

Aiming at prediction of telecom customer churn,a novel method was proposed to increase the prediction accuracy with the missing data based on the Bayesian network.This method used k-nearest neighbor algorithm to fill the missing data and adds two types of monotonic influence constraints into the process of learning Bayesian network parameter.Simulations and actual data analysis demonstrate that the proposed algorithm obtains higher prediction accuracy of churn customers with the loss of less cost prediction accuracy of loyal customers,outperforms the classic expectation maximization algorithm.

Key words: Bayesian network, parameter learning, data missing, nearest neighbor algorithm, qualitative constraint

CLC Number:

TP181

Yuxiang ZHAO,Guangyue LU,Hanglong WANG,Siwei LI. A prediction algorithm of telecom customer churn based on Bayesian network parameters learning under incomplete data[J]. Telecommunications Science, 2018, 34(1): 52-60.

Figures/Tables 10

References 16

[1]	JAMIL S , KHAN A . Churn comprehension analysis for telecommunication industry using ALBA[C]// International Conference on Emerging Technologies(ICET 2016),Oct 1,2016,Islamabad,Pakistan. Piscataway:IEEE Press, 2016.
[2]	PEARL J . Probabilistic reasoning in intelligent systems:networks of plausible inference[J]. Computer Science Artificial Intelligence, 1988,70(2): 1022-1027.
[3]	WAGNER S . A Bayesian network approach to assess and predict software quality using activity-based quality models[J]. Information and Software Technology, 2016,52(11): 1230-1241.
[4]	PENDHARKAR P C , KHOSROWPOUR M , RODGER J A . Application of Bayesian network classifiers and data envelopment analysis for mining breast cancer patterns[J]. Journal of Computer Information Systems, 2016,40(4): 127-131.
[5]	COOK J , LEWANDOWSKY S . Rational irrationality:modeling climate change belief polarization using Bayesian networks[J]. Topics in Cognitive Science, 2016,8(1): 160-179.
[6]	LAURITZEN S L . The EM algorithm for graphical association models with missing data[J]. Computational Statistics ＆ Data Analysis, 1995,19(2): 191-201.
[7]	LIAO W , JI Q . Learning Bayesian network parameters under incomplete data with domain knowledge[J]. Pattern Recognition, 2009,42(11): 3046-3056.
[8]	MASEGOSA A R , FEELDERS A J , GAAG L C V D . Learning from incomplete data in Bayesian networks with qualitative influences[J]. International Journal of Approximate Reasoning, 2015,69(C): 18-34.
[9]	CORANI G , CAMPOS C P D . A maximum entropy approach to learn Bayesian networks from incomplete data[M]// Interdisciplinary Bayesian Statistics. Berlin: Springer International Publishing, 2015: 69-82.
[10]	张连文 . 贝叶斯网引论[M]. 北京: 科学出版社, 2006.
	Zhang L W . Introduction of Bayesian network[M]. Beijing: Science Press, 2006.
[11]	ZHOU Y , FENTON N , ZHU C . An empirical study of Bayesian network parameter learning with monotonic influence constraints[J]. Decision Support Systems, 2016,87(C): 69-79.
[12]	FEELDERS A , VAN D G L C . Learning Bayesian network parameters under order constraints[J]. International Journal of Approximate Reasoning, 2005,42(1-2): 37-53.
[13]	刘星毅 . 基于马氏距离和灰色分析的缺失值填充算法[J]. 计算机应用, 2009,29(9): 2502-2504.
	LIU X Y . Improved kNN algorithm based on Mahalanobis distance and gray analysis[J]. Journal of Computer Applications, 2009,29(9): 2502-2504.
[14]	杨宇, 高晓光, 郭志高 . 小数据集条件下基于数据再利用的BN参数学习[J]. 自动化学报, 2015,41(12): 2058-2071.
	YANG Y , GAO X G , GUO Z G . Learning BN parameters with small data sets based reutilization[J]. Acta Automatica Sinica, 2015,41(12): 2058-2071.
[15]	KULLBACK S , LEIBLER R A . On Information and Sufficiency[J]. Annals of Mathematical Statistics, 1951,22(22): 79-86.
[16]	HU X J , YANG S L , MA X J . Inference structure and construction algorithms of Bayesian network based on junction tree[J]. Acta Simulata Systematica Sinica, 2004,16(11): 2559-2558.

Metrics

Recommended 0

No Suggested Reading articles found!

内部约束	外部约束
P (S=1\|C=0)>P(S=0\|C=0)	P (S=1\|C=0)>P(S=1\|C=1)
P (S=0\|C=1)>P(S=1\|C=1)	P (R=1\|C=1)>P(R=1\|C=0)
P (R=1\|C=1)>P(R=0\|C=1)	P (W =1\|S=1,R=1)>P(W =1\|S=1,R=0)
P (R=0\|C=0)>P(R=1\|C=0)	P (W =1\|S=1,R=1)>P(W =1\|S=0,R=1)
P (W =1\|S=1,R=1)>P(W =0\|S=1,R=1)	P (W =1\|S=1,R=1)>P(W =1\|S=0,R=0)
P (W =1\|S=0,R=1)>P(W =0\|S=0,R=1)	P (W =1\|S=0,R=1)>P(W =1\|S=0,R=0)
P (W =1\|S=1,R=0)>P(W =0\|S=1,R=0)	P (W =1\|S=1,R=0)>P(W =1\|S=0,R=0)

约束类别	约束条件
内部约束	P (D=1\|B=1,C=2)>P(D=2\|B=1,C=2)
	P(D=1\|B=2,C=2)>P(D=2\|B=2,C=2)
	P(E=3\|C=2,D=1)>P(E=2\|C=2,D=1)>P(E=1\|C=2,D=1)
	P(F=1\|E=1)>P(F=2\|E=1)>P(F=3\|E=1)
	P(H =1\|E=1)>P(H=2\|E=1)>P(H =3\|E=1)
	P(G=1\|D=1,E=1)>P(G=2\|D=1,E=1)>P(G=3\|D=1,E=1)
	P(G=3\|D=1,E=3)>P(G=2\|D=1,E=3)>P(G=1\|D=1,E=3)
外部约束	P (C=2\|A=3)>P(C=2\|A=2)>P(C=2\|A=1)
	P(D=1\|B=2,C=2)>P(D=1\|B=1,C=2)>P(D=1\|B=2,C=1)>P(D=1\|B=1,C=1)
	P (F=1\|E=1)>P(F=1\|E=2)>P(F=1\|E=3)
	P(I=1\|D=1,H =3)>P(I=1\|D=1,H =2)>P(I=1\|D=1,H =1)
	P(G=1\|D=1,E=1)>P(G=1\|D=1,E=2)>P(G=1\|D=1,E=3)
	P(G=3\|D=1,E=3)>P(G=3\|D=1,E=2)>P(G=3\|D=1,E=1)

电信数据	kNN-DCE	EM	qirEM
SE	84.56%	70.49%	75.4%
SP	82%	90%	88%
Gm	83.27%	79.37%	81.45%

A prediction algorithm of telecom customer churn based on Bayesian network parameters learning under incomplete data

RichHTML

PDF下载

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 10

References 16

Related Articles 2

Metrics

Recommended 0

[1]	Hao Wang,Yunfei Lv,Yuanbao Chen,Yunfei Peng. Predicting Large-Scale WLAN Traffic via Granger Causality Based Bayesian Network [J]. Telecommunications Science, 2015, 31(8): 46-50.
[2]	Hongxin Gui,Changhao Yang,Fei Cheng. Research on Customer Loss Prediction for Mobile Service Based on Bayesian Network [J]. Telecommunications Science, 2009, 25(3): 70-75.