基于SHAP解释方法的智慧居家养老服务平台用户流失预测研究<sup>*</sup>

doi:10.11925/infotech.2096-3467.2022.1168

数据分析与知识发现

2024, Vol. 8

Issue (1): 40-54 https://doi.org/10.11925/infotech.2096-3467.2022.1168

研究论文

本期目录 | 过刊浏览 | 高级检索

基于SHAP解释方法的智慧居家养老服务平台用户流失预测研究^*

刘天畅(

),王雷,朱庆华

南京大学信息管理学院南京 210023

Predicting User Churn of Smart Home-based Care Services Based on SHAP Interpretation

Liu Tianchang(

),Wang Lei,Zhu Qinghua

School of Information Management, Nanjing University, Nanjing 210023, China

摘要
图/表
参考文献
相关文章
Metrics

全文: PDF (1914 KB) HTML ( 19 )
输出: BibTeX | EndNote (RIS)

摘要

【目的】 构建智慧居家养老服务平台用户流失预测模型，并使用SHAP解释方法分析不同特征的影响。【方法】 基于智慧居家养老服务平台用户在2019年至2021年三年间产生的超过30万条社区居家养老服务订单数据，通过改进的RFM模型（RFM-MLP）、马斯洛需求层次理论、安德森模型并结合Boruta算法确定用户价值特征、服务选择特征、个人特征三类共11个特征。建立5种机器学习模型，从中选择效果最好的XGBoost模型预测用户流失，运用SHAP解释方法完成特征影响全局解释、特征依赖分析、单样本解释分析。【结果】 模型预测结果准确率和F1值均达到87%左右，家政服务服务购买次数、留存天数、年龄等是预测养老服务平台用户流失的重要特征。【局限】 仅选取一个地区的数据进行分析，数据量和算法复杂度方面还有提升空间。【结论】 SHAP解释方法可以兼顾机器学习预测模型的精度和解释性，能够为智慧居家养老服务平台在运营策略和内容设计方面的优化提供依据。

	服务

	把本文推荐给朋友
	加入引用管理器
	E-mail Alert
	RSS
	作者相关文章
	刘天畅
	王雷
	朱庆华

关键词 ：智慧养老, 用户流失, XGBoost, 可解释性机器学习, SHAP

Abstract：

[Objective] This study constructs a user churn prediction model for smart home-based care services. It utilizes the SHAP interpretation method to analyze the impact of different features on user churn. [Methods] First, we retrieved more than 300,000 community home-based care service orders from 2019 to 2021. Then, we incorporated the RFM model (RFM-MLP), the Maslow’s hierarchy of demand theory, the Anderson model, and the Boruta algorithm to identify 11 characteristics across three categories: user values, service selections, and individual features. Third, we chose the XGBoost model from the five established machine learning models for the best performance in predicting user churn. Finally, we employed the SHAP interpretation method to examine the feature impact, dependence, and single-sample analysis. [Results] The predictive model achieves high accuracy and F1 score of approximately 87%. Noteworthy features for predicting user churn on smart home-based care services include domestic service purchase numbers, use length, and user age. [Limitations] Our data was from a single region. The data quality and algorithm complexity could be improved in the future. [Conclusions] The SHAP interpretation method effectively balances accuracy and interpretability in machine learning prediction models. The insights gained provide a foundation for optimizing operational strategies and content design on smart home-based care service platforms.

Key words： Smart Aging User Churn XGBoost Interpretable Machine Learning SHAP

收稿日期: 2022-11-06 出版日期: 2023-03-22

ZTFLH:	TP393
	G250

基金资助:*国家社会科学基金重大项目(22&ZD327);江苏省高校哲学社会科学研究重大项目(2021SJZDA044)

通讯作者: 刘天畅，ORCID：0000-0002-1381-3559，E-mail：njutcl@smail.nju.edu.cn。

引用本文:

刘天畅, 王雷, 朱庆华. 基于SHAP解释方法的智慧居家养老服务平台用户流失预测研究^*[J]. 数据分析与知识发现, 2024, 8(1): 40-54.
Liu Tianchang, Wang Lei, Zhu Qinghua. Predicting User Churn of Smart Home-based Care Services Based on SHAP Interpretation. Data Analysis and Knowledge Discovery, 2024, 8(1): 40-54.

链接本文:

https://manu44.magtech.com.cn/Jwk_infotech_wk3/CN/10.11925/infotech.2096-3467.2022.1168 或 https://manu44.magtech.com.cn/Jwk_infotech_wk3/CN/Y2024/V8/I1/40

Fig.1 研究框架

Fig.2 用户流失特征理论体系

Table 1 特征汇总表

Fig.3 流失期限与回购率示意图

Table 2 特征选择结果

Table 3 特征描述性分析

Table 4 模型性能对比

Table 5 不同模型性能差异显著性

Fig.4 特征重要性排名和特征对模型预测的影响

Table 6 逻辑回归结果

Fig.5 用户价值特征SHAP值散点图

Fig.6 服务选择特征SHAP值散点图

Fig.7 个人特征SHAP值散点图

Fig.8 两位老年用户的单样本解释力图

[1]	国务院. 国务院关于加快发展养老服务业的若干意见[EB/OL]. (2013-09-13). [2022-02-28]. http://www.gov.cn/zwgk/2013-09/13/content_2487704.htm.
[1]	(The State Council. Several Opinions of the State Council on Accelerating the Development of the Elderly Care Service Industry[EB/OL]. (2013-09-13). [2022-02-28]. http://www.gov.cn/zwgk/2013-09/13/content_2487704.htm.)
[2]	曾起艳, 何志鹏, 曾寅初. 老年人居家养老服务需求意愿与行为悖离的原因分析[J]. 人口与经济, 2022(2): 87-103.
[2]	(Zeng Qiyan, He Zhipeng, Zeng Yinchu. The Cause of Paradoxical Between Willingness and Behavior of Elderly People’s Demand for Home-Based Care Services[J]. Population & Economics, 2022(2): 87-103.)
[3]	白玫, 朱庆华. 老年用户智慧养老服务需求及志愿服务意愿影响因素分析——以武汉市江汉区为例[J]. 现代情报, 2018, 38(12): 3-8. doi: 10.3969/j.issn.1008-0821.2018.12.001
[3]	(Bai Mei, Zhu Qinghua. Impact Factors of Smart Care Needs and Volunteer Service Willingness for the Aged——A Case of Jianghan District in Wuhan[J]. Journal of Modern Information, 2018, 38(12): 3-8.) doi: 10.3969/j.issn.1008-0821.2018.12.001
[4]	冯春梅. 新型社区居家养老服务的影响因素分析[J]. 统计与决策, 2018, 34(20): 110-113.
[4]	(Feng Chunmei. Analysis on Influencing Factors of New Community Home Care Service for the Aged[J]. Statistics & Decision, 2018, 34(20): 110-113.)
[5]	Shrestha Y R, He V F, Puranam P, et al. Algorithm Supported Induction for Building Theory: How Can We Use Prediction Models to Theorize?[J]. Organization Science, 2021, 32(3): 856-880. doi: 10.1287/orsc.2020.1382
[6]	Molnar C, Casalicchio G, Bischl B. Interpretable Machine Learning - A Brief History, State-of-the-Art and Challenges[C]// Proceedings of the 2020 Workshops of the European Conference on Machine Learning and Knowledge Discovery in Databases. Cham: Springer, 2020: 417-431.
[7]	徐孝娟, 赵宇翔, 朱庆华. 社交网站用户流失行为理论基础及影响因素探究[J]. 图书情报工作, 2016, 60(4): 134-141. doi: 10.13266/j.issn.0252-3116.2016.04.018
[7]	(Xu Xiaojuan, Zhao Yuxiang, Zhu Qinghua. Theoretical Basis and Influence Factors of User Exodus Behavior of Social Networking Sites[J]. Library and Information Service, 2016, 60(4): 134-141.) doi: 10.13266/j.issn.0252-3116.2016.04.018
[8]	陈渝, 黄亮峰. 理性选择理论视角下的电子书阅读客户端用户流失行为研究[J]. 图书馆论坛, 2019, 39(9): 118-126.
[8]	(Chen Yu, Huang Liangfeng. Empirical Research on Customers’ Churn Behavior on E-Book Reading Apps: Based on Rational Choice Theory[J]. Library Tribune, 2019, 39(9): 118-126.)
[9]	徐孝娟, 赵宇翔, 吴曼丽, 等. S-O-R理论视角下的社交网站用户流失行为实证研究[J]. 情报杂志, 2017, 36(7): 188-194.
[9]	(Xu Xiaojuan, Zhao Yuxiang, Wu Manli, et al. The Empirical Research of User Exodus in Social Network Based on the Stimuli-Organism-Response Theory[J]. Journal of Intelligence, 2017, 36(7): 188-194.)
[10]	郭顺利, 张向先, 相甍甍. 高校图书馆微信公众平台用户流失行为模型及其影响因素分析[J]. 图书情报工作, 2017, 61(2): 57-66. doi: 10.13266/j.issn.0252-3116.2017.02.007
[10]	(Guo Shunli, Zhang Xiangxian, Xiang Mengmeng. Research on the Customer Churn Behavior Model and Its Influencing Factors of WeChat Public Platform in University Libraries[J]. Library and Information Service, 2017, 61(2): 57-66.) doi: 10.13266/j.issn.0252-3116.2017.02.007
[11]	郑德俊, 李杨, 沈军威, 等. 移动阅读服务平台的用户流失因素分析——以“微信读书”平台为例[J]. 情报理论与实践, 2019, 42(8): 78-82.
[11]	(Zheng Dejun, Li Yang, Shen Junwei, et al. A Study on the Influence Factors of User Exodus in Mobile Reading Platform: Taking “WeChat Reading” as an Example[J]. Information Studies: Theory & Application, 2019, 42(8): 78-82.)
[12]	王锰, 华钰文, 陈雅. S-O-R理论视角下东部地区乡村公共数字文化服务用户流失行为研究[J]. 图书馆杂志, 2022, 41(2): 36-46.
[12]	(Wang Meng, Hua Yuwen, Chen Ya. A Study on User Churn of Rural Public Digital Cultural Services in Eastern China from the Perspective of S-O-R Theory[J]. Library Journal, 2022, 41(2): 36-46.)
[13]	邢绍艳, 朱学芳. 付费知识直播用户流失预测实证研究[J]. 信息资源管理学报, 2022, 12(4): 121-130.
[13]	(Xing Shaoyan, Zhu Xuefang. An Empirical Study on the User Churn Prediction of Paid Knowledge Live[J]. Journal of Information Resources Management, 2022, 12(4): 121-130.)
[14]	Tarokh M J, EsmaeiliGookeh M. Modeling Patient’s Value Using a Stochastic Approach: An Empirical Study in the Medical Industry[J]. Computer Methods and Programs in Biomedicine, 2019, 176: 51-59. doi: 10.1016/j.cmpb.2019.04.021
[15]	Sato K, Oka M, Kato K. Early Churn User Classification in Social Networking Service Using Attention-Based Long Short-Term Memory[C]// Proceedings of the 2019 Pacific-Asia Conference on Knowledge Discovery and Data Mining. Cham: Springer, 2019: 45-56.
[16]	Kostić S M, Simić M I, Kostić M V. Social Network Analysis and Churn Prediction in Telecommunications Using Graph Theory[J]. Entropy, 2020, 22(7): Article No.753.
[17]	Kilimci Z H, Yörük H, Akyokus S. Sentiment Analysis Based Churn Prediction in Mobile Games Using Word Embedding Models and Deep Learning Algorithms[C]// Proceedings of the 2020 International Conference on Innovations in Intelligent Systems and Applications. IEEE, 2020: 1-7.
[18]	冯鑫, 王晨, 刘苑, 等. 基于评论情感倾向和神经网络的客户流失预测研究[J]. 中国电子科学研究院学报, 2018, 13(3): 340-345.
[18]	(Feng Xin, Wang Chen, Liu Yuan, et al. The Customer Churn Prediction Based on Emotional Polarity and BPNN[J]. Journal of China Academy of Electronics and Information Technology, 2018, 13(3): 340-345.)
[19]	朱雅彬. 高校移动图书馆App用户流失实证研究[J]. 图书馆学研究, 2020(10): 39-45.
[19]	(Zhu Yabin. An Empirical Research of User Churn in University Mobile Library App[J]. Research on Library Science, 2020(10): 39-45.)
[20]	王若佳, 严承希, 郭凤英, 等. 基于用户画像的在线健康社区用户流失预测研究[J]. 数据分析与知识发现, 2022, 6(2/3): 80-92.
[20]	(Wang Ruojia, Yan Chengxi, Guo Fengying, et al. Predicting Churners of Online Health Communities Based on the User Persona[J]. Data Analysis and Knowledge Discovery, 2022, 6(2/3): 80-92.)
[21]	Lundberg S M, Lee S I. A Unified Approach to Interpreting Model Predictions[C]// Proceedings of the 31st International Conference on Neural Information Processing Systems. ACM, 2017: 4768-4777.
[22]	Shapley L. A Value for n-Person Games[A]//Kuhn H W. Classics in Game Theory[M]. Princeton University Press, 1997: 69-79.
[23]	Lundberg S M, Erion G, Chen H, et al. Explainable AI for Trees: From Local Explanations to Global Understanding[OL]. arXiv Preprint, arXiv: 1905.04610.
[24]	雷欣南, 林乐凡, 肖斌卿, 等. 小微企业违约特征再探索:基于SHAP解释方法的机器学习模型[J/OL]. 中国管理科学. https://doi.org/10.16381/j.cnki.issn1003-207x.2021.0027.
[24]	(Lei Xinnan, Lin Lefan, Xiao Binqing, et al. Re-Exploration of Small and Micro Enterprises’ Default Characteristics Based on Machine Learning Models with SHAP[J/OL]. Chinese Journal of Management Science. https://doi.org/10.16381/j.cnki.issn1003-207x.2021.0027.)
[25]	廖彬, 王志宁, 李敏, 等. 融合XGBoost与SHAP模型的足球运动员身价预测及特征分析方法[J]. 计算机科学, 2022, 49(12): 195-204. doi: 10.11896/jsjkx.210600029
[25]	(Liao Bin, Wang Zhining, Li Min, et al. Integrating XGBoost and SHAP Model for Football Player Value Prediction and Characteristic Analysis[J]. Computer Science, 2022, 49(12): 195-204.) doi: 10.11896/jsjkx.210600029
[26]	曹睿, 廖彬, 李敏, 等. 基于XGBoost的在线短租市场价格预测及特征分析模型[J]. 数据分析与知识发现, 2021, 5(6): 51-65.
[26]	(Cao Rui, Liao Bin, Li Min, et al. Predicting Prices and Analyzing Features of Online Short-Term Rentals Based on XGBoost[J]. Data Analysis and Knowledge Discovery, 2021, 5(6): 51-65.)
[27]	李宗敏, 张琪, 杜鑫雨. 基于辟谣微博的互动及热门评论情感倾向的辟谣效果研究——以新冠疫情相关辟谣微博为例[J]. 情报杂志, 2020, 39(11): 90-95.
[27]	(Li Zongmin, Zhang Qi, Du Xinyu. Research on Rumor-Refutation Effectiveness Based on the Interactions and Popular Comments’ Emotional Tendencies of the Rumor-Refuting Microblogs: Taking Rumor-Refuting Microblogs Related with COVID-2019 as an Example[J]. Journal of Intelligence, 2020, 39(11): 90-95.)
[28]	Parsa A B, Movahedi A, Taghipour H, et al. Toward Safer Highways, Application of XGBoost and SHAP for Real-Time Accident Detection and Feature Analysis[J]. Accident Analysis & Prevention, 2020, 136: Article No.105405.
[29]	Meng Y, Yang N H, Qian Z L, et al. What Makes an Online Review More Helpful: An Interpretation Framework Using XGBoost and SHAP Values[J]. Journal of Theoretical and Applied Electronic Commerce Research, 2020, 16(3): 466-490. doi: 10.3390/jtaer16030029
[30]	卢云, 张梦月, 夏赫, 等. 基于LightGBM及SHAP对1055例新型冠状病毒肺炎重型患者中西医结合及西医治疗的多中心回顾性研究[J]. 北京中医药大学学报, 2021, 44(12): 1098-1107.
[30]	(Lu Yun, Zhang Mengyue, Xia He, et al. Multicenter Retrospective Analysis of 1055 Severe Cases of COVID-19 Treated by Integrated Chinese and Western Medicine or Western Medicine Based on LightGBM and SHAP[J]. Journal of Beijing University of Traditional Chinese Medicine, 2021, 44(12): 1098-1107.)
[31]	Wen X, Xie Y C, Wu L T, et al. Quantifying and Comparing the Effects of Key Risk Factors on Various Types of Roadway Segment Crashes with LightGBM and SHAP[J]. Accident Analysis & Prevention, 2021, 159: Article No.106261.
[32]	丁恒, 阮靖龙. 基于算法归因框架的LIS领域学者施引影响因素实证研究[J]. 图书情报知识, 2022, 39(2): 83-97.
[32]	(Ding Heng, Ruan Jinglong. Exploring the Factors Influencing LIS Scholars Citing Other’s Works: An Empirical Research Based on Algorithmic Attribution[J]. Documentation, Information & Knowledge, 2022, 39(2): 83-97.)
[33]	Hughes A M. Strategic Datebase Marketing[M]. Chicago: Probus Publishing, 1994.
[34]	Cheng C H, Chen Y S. Classifying the Segmentation of Customer Value via RFM Model and RS Theory[J]. Expert Systems with Applications, 2009, 36(3): 4176-4184. doi: 10.1016/j.eswa.2008.04.003
[35]	王天慧. 基于RFM改进模型的客户关系管理系统客户分类研究与应用[D]. 重庆: 重庆大学, 2019.
[35]	(Wang Tianhui. Customer Segmentation Study of CRM Based on Improved RFM Model[D]. Chongqing: Chongqing University, 2019.)
[36]	魏玲, 郭新悦. 基于改进 RFM与 GMDH算法的MOOC用户流失预测[J]. 中国远程教育, 2020(9): 39-43, 61.
[36]	(Wei Ling, Guo Xinyue. Using Adapted RFM and GMDH Algorithms to Predict MOOC User Attrition Rate[J]. Chinese Journal of Distance Education, 2020(9): 39-43, 61.)
[37]	Keaveney S M. Customer Switching Behavior in Service Industries: An Exploratory Study[J]. Journal of Marketing, 1995, 59(2): 71-82.
[38]	Maslow A H. A Theory of Human Motivation[J]. Psychological Review, 1943, 50(4): 370-396. doi: 10.1037/h0054346
[39]	侯冰. 老年人社区居家养老服务需求层次及其满足策略研究[J]. 社会保障评论, 2019, 3(3): 147-159.
[39]	(Hou Bing. Community Home-Based Care Service for Urban Elderly: Demand Levels and Satisfying Strategy[J]. Chinese Social Security Review, 2019, 3(3): 147-159.)
[40]	李斌, 王依明, 李雪, 等. 城市社区养老服务需求及其影响因素[J]. 建筑学报, 2016(S1): 90-94.
[40]	(Li Bin, Wang Yiming, Li Xue, et al. The Need and Influence Factors of the Elderly Care Services in Urban Community[J]. Architectural Journal, 2016(S1): 90-94.)
[41]	Kominski G F. Changing the US Health Care System: Key Issues in Health Services Policy and Management[M]. John Wiley & Sons, 2013.
[42]	李月娥, 卢珊. 医疗卫生领域安德森模型的发展、应用及启示[J]. 中国卫生政策研究, 2017, 10(11): 77-82.
[42]	(Li Yue’e, Lu Shan. The Development, Application and Implications of the Anderson Model in the Field of Healthcare[J]. Chinese Journal of Health Policy, 2017, 10(11): 77-82.)
[43]	彭希哲, 宋靓珺, 黄剑焜. 中国失能老人长期照护服务使用的影响因素分析——基于安德森健康行为模型的实证研究[J]. 人口研究, 2017, 41(4): 46-59.
[43]	(Peng Xizhe, Song Liangjun, Huang Jiankun. Determinants of Long-Term Care Services Among Disabled Older Adults in China: A Quantitative Study Based on Andersen’s Behavioral Model[J]. Population Research, 2017, 41(4): 46-59.)
[44]	Hu B, Li B Q, Wang J, et al. Home and Community Care for Older People in Urban China: Receipt of Services and Sources of Payment[J]. Health & Social Care in the Community, 2020, 28(1): 225-235.
[45]	周婉婷, 赵志杰, 刘阳, 等. 电子商务客户流失的DBN预测模型研究[J]. 计算机工程与应用, 2022, 58(11): 84-92. doi: 10.3778/j.issn.1002-8331.2104-0221
[45]	(Zhou Wanting, Zhao Zhijie, Liu Yang, et al. Research on DBN Prediction Model of E-Commerce Customer Churn[J]. Computer Engineering and Applications, 2022, 58(11): 84-92.) doi: 10.3778/j.issn.1002-8331.2104-0221
[46]	Chawla N V, Bowyer K W, Hall L O, et al. SMOTE: Synthetic Minority Over-Sampling Technique[J]. Journal of Artificial Intelligence Research, 2002, 16: 321-357. doi: 10.1613/jair.953
[47]	Bambi C, Modesto L. Rotating Regular Black Holes[J]. Physics Letters B, 2013, 721(4-5): 329-334. doi: 10.1016/j.physletb.2013.03.025
[48]	Kursa M B, Rudnicki W R. Feature Selection with the Boruta Package[J]. Journal of Statistical Software, 2010, 36(11): 1-13.
[49]	Dietterich T G. Approximate Statistical Tests for Comparing Supervised Classification Learning Algorithms[J]. Neural Computation, 1998, 10(7): 1895-1923. doi: 10.1162/089976698300017197 pmid: 9744903
[50]	郭丽娜, 郝勇. 个体健康、家庭照护和社会供给:谁更影响老人的居家养老服务需求[J]. 西北人口, 2019, 40(5): 36-49.
[50]	(Guo Lina, Hao Yong. Health Conditions, Informal Care and Social Provision: Which is More Influencing the Elderly’s Demand for Home-Care[J]. Northwest Population Journal, 2019, 40(5): 36-49.)
[51]	阳义南, 袁涛. 养老服务购买者的甄别与归因分解[J]. 中国人口科学, 2022(1): 113-125.
[51]	(Yang Yinan, Yuan Tao. Analysis on Identification of the Elderly Care Service Buyers and the Attribution Decomposition[J]. Chinese Journal of Population Science, 2022(1): 113-125.)
[52]	王琼. 城市社区居家养老服务需求及其影响因素——基于全国性的城市老年人口调查数据[J]. 人口研究, 2016, 40(1): 98-112.
[52]	(Wang Qiong. Demands and Determinants of Community Home-Based Care Services for Urban Elderly: Based on the 2010 National Elderly Survey in China[J]. Population Research, 2016, 40(1): 98-112.)

[1]	刘智锋, 王继民. 可解释机器学习在信息资源管理领域的应用研究综述^*[J]. 数据分析与知识发现, 2024, 8(1): 16-29.
[2]	王若佳, 严承希, 郭凤英, 王继民. 基于用户画像的在线健康社区用户流失预测研究^*[J]. 数据分析与知识发现, 2022, 6(2/3): 80-92.
[3]	王琰, 胥美美, 童俞嘉, 苟欢, 蔡荣, 单治易, 安新颖. 基于机器学习的环境监测数据对循环系统疾病死亡影响及预测预警模型构建^*[J]. 数据分析与知识发现, 2022, 6(10): 79-92.
[4]	刘渊晨, 王昊, 高亚琪. 在线音乐歌单播放量预测及影响因素分析^*[J]. 数据分析与知识发现, 2021, 5(8): 100-112.
[5]	曹睿,廖彬,李敏,孙瑞娜. 基于XGBoost的在线短租市场价格预测及特征分析模型^*[J]. 数据分析与知识发现, 2021, 5(6): 51-65.
[6]	丁勇,陈夕,蒋翠清,王钊. 一种融合网络表示学习与XGBoost的评分预测模型*[J]. 数据分析与知识发现, 2020, 4(11): 52-62.
[7]	李晓峰,马静,李驰,朱恒民. 基于XGBoost模型的电商商品品名识别算法研究 ^*[J]. 数据分析与知识发现, 2019, 3(7): 34-41.
[8]	杨贵军,徐雪,赵富强. 基于XGBoost算法的用户评分预测模型及应用^*[J]. 数据分析与知识发现, 2019, 3(1): 118-126.

Viewed

Full text

Abstract

Cited

Shared

Discussed