logo

SCIENCE CHINA Information Sciences, Volume 62, Issue 5: 052101(2019) https://doi.org/10.1007/s11432-018-9511-1

A trust-aware random walk model for return propensity estimation and consumer anomaly scoring in online shopping

More info
  • ReceivedApr 23, 2018
  • AcceptedJun 25, 2018
  • PublishedMar 19, 2019

Abstract

In online shopping, most of consumers will not clear their return reasons when submitting return requests (e.g., select the option “other reasons").Prior literature mostly investigates into the return event at the transaction level, and the underlying force of returns remains untracked.To deal with this problem, we propose a machine learning algorithm named as trust-aware random walk model (TARW).In the proposed model, four patterns of consumers can be identified in terms of return forces: (i) selfish consumers, (ii) honest consumers, (iii) fraud consumers, and (iv) irrelevant consumers.To profile consumers' return patterns, we capture consumers' similarities in order preferences and return tendencies separately.Based on consumers' similarities, we obtain a return pattern trust network by introducing the trust network and collaborative filtering algorithms.Subsequently, we develop two important applications based on the trust network:(i) estimating consumers' return propensities for product types;(ii) scoring the anomaly for consumers' returns for one product.Finally, we conduct extensive experiments with the real-world data to validate the model's effectiveness in predicting and tracing consumers' returns.With the proposed model, we can help retailers improve the conversion rates of selfish consumers, retain honest consumers, and block fraud consumers.


Acknowledgment

This work was supported by National Key RD Program of China (Grant No. 2018YFB-1004300), National Natural Science Foundation of China (Grant Nos. 61773199, 71732002), and Philosophy and Social Science Foundation of Higher Education Institutions of Jiangsu Province, China (Grant No. 2017SJB0006).


References

[1] Fu Y, Liu G, Papadimitriou S. Fused latent models for assessing product return propensity in online commerce. Decision Support Syst, 2016, 91: 77-88 CrossRef Google Scholar

[2] Dutta S, Biswas A, Grewal D. Regret from postpurchase discovery of lower market prices: do price refunds help? J Marketing, 2011, 75: 124--138. Google Scholar

[3] Reason?code based model to forecast product returns. Foresight, 2012, 14: 105-120 CrossRef Google Scholar

[4] Wood S L. Remote Purchase Environments: The Influence of Return Policy Leniency on Two-Stage Decision Processes. J Marketing Res, 2001, 38: 157-169 CrossRef Google Scholar

[5] Ang L, Dubelaar C, Lee B C. To trust or not to trust? A model of internet trust from the customer's point of view. In: Proceedings of BLED 2001, Bled Austria, 2001. 43. Google Scholar

[6] Zacharia G, Moukas A, Maes P. Collaborative reputation mechanisms for electronic marketplaces. Decision Support Syst, 2000, 29: 371-388 CrossRef Google Scholar

[7] Dellarocas C. The Digitization of Word of Mouth: Promise and Challenges of Online Feedback Mechanisms. Manage Sci, 2003, 49: 1407-1424 CrossRef Google Scholar

[8] Zhang Y, Liu T, Li R, et al. Evaluation model of buyers' dynamic reputation in e-commerce. Int J Multimedia Ubiquitous Eng, 2015, 10: 53-64. Google Scholar

[9] Grace A M, Williams S O. Comparative Analysis of Neural Network and Fuzzy Logic Techniques in Credit Risk Evaluation. Int J Intelligent Inf Technologies, 2016, 12: 47-62 CrossRef Google Scholar

[10] Gunn S R. Support vector machines for classification and regression. Tech Rep, 1998, 14: 5-16. Google Scholar

[11] Lee T S, Chiu C C, Lu C J. Credit scoring using the hybrid neural discriminant technique. Expert Syst Appl, 2002, 23: 245-254 CrossRef Google Scholar

[12] Wang Q, Zhang Y Y, Dai X Y. A high strength pH responsive supramolecular copolymer hydrogel. Sci China Technol Sci, 2017, 60: 78-83 CrossRef Google Scholar

[13] Jiang J, Li Y J, Feng Q Y. A multiple user sharing behaviors based approach for fake file detection in P2P environments. Sci China Inf Sci, 2010, 53: 2169-2184 CrossRef Google Scholar

[14] MassaP A. Trust-aware collaborative filtering for recommender systems. In: Proceedings of International Conference on Cooperative Information Systems. Berlin: Springer Press, 2004. 492--508. Google Scholar

[15] Sarwar B, Karypis G, Konstan J, et al. Item-based collaborative filtering recommendation algorithms. In: Proceedings of the 10th International Conference on World Wide Web. Raleigh: ACM Press, 2001. 285--295. Google Scholar

[16] Shepitsen A, Gemmell J, Mobasher B, et al. Personalized recommendation in social tagging systems using hierarchical clustering. In: Proceedings of the 2008 ACM Conference on Recommender Systems. Lausanne: ACM Press, 2008. 259--266. Google Scholar

[17] Xiwang Yang , Yang Guo , Yong Liu . Bayesian-Inference-Based Recommendation in Online Social Networks. IEEE Trans Parallel Distrib Syst, 2013, 24: 642-651 CrossRef Google Scholar

[18] Qian Y, Zhiyong P, Liang H, et al. A latent topic based collaborative filtering recommendation algorithm for web communities. In: Proceedings of 2012 Web Information Systems and Applications Conference. Hainan: IEEE Press, 2012. 241--246. Google Scholar

[19] Chen D E, Ying Y L. A collaborative filtering recommendation algorithm based on bipartite graph. Advanced Materials Research, 2013, 756: 3865--3868. Google Scholar

[20] Rong H G, Zhou X, Yang C, et al. The rich and the poor: a Markov decision process approach to optimizing taxi driver revenue efficiency. In: Proceedings of the 25th ACM International Conference on Information and Knowledge Management (CIKM). Indianapolis: ACM Press, 2016. 2329--2334. Google Scholar

[21] Yu W. Analysis on trust influencing factors and trust model from multiple perspectives of online Auction. Cent Eur J Phys, 2017, 15: 613--619. Google Scholar

[22] Fouss F, Pirotte A, Renders J. Random-Walk Computation of Similarities between Nodes of a Graph with Application to Collaborative Recommendation. IEEE Trans Knowl Data Eng, 2007, 19: 355-369 CrossRef Google Scholar

[23] Haveliwala T H. Topic-sensitive pagerank: A context-sensitive ranking algorithm for web search. IEEE Trans Knowl Data Eng, 2003, 15: 784-796 CrossRef Google Scholar

[24] Sun J, Qu H, Chakrabarti D, et al. Neighborhood formation and anomaly detection in bipartite graphs. In: Proceedings of the 5th IEEE International Conference on Data Mining. Houston: IEEE Press, 2005. 8. Google Scholar

[25] Jamali M, Ester M. Trustwalker: a random walk model for combining trust-based and item-based recommendation. In: Proceedings of the 5th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. Paris: ACM Press, 2009. 397--406. Google Scholar

[26] Jeh G, Widom J. SimRank: a measure of structural-context similarity. In: Proceedings of the 8th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. Edmonton: ACM Press, 2002. 538--543. Google Scholar

[27] Zhang Z, Zeng D D, Abbasi A, et al. A random walk model for item recommendation in social tagging systems. ACM Transactions on Management Information Systems (TMIS), 2013, 4: 1-24. Google Scholar

[28] Alexandridis G, Siolas G, Stafylopatis A. Accuracy versus novelty and diversity in recommender systems: a nonuniform random walk approach. In: Recommendation and Search in Social Networks. Berlin: Springer Press, 2015. 41--57. Google Scholar

[29] Gong J, Gao X, Cheng H, et al. Integrating a weighted-average method into the random walk framework to generate individual friend recommendations. Sci China Inf Sci, 2017, 60: 1-22. Google Scholar

[30] Smith W R. Product Differentiation and Market Segmentation as Alternative Marketing Strategies. J Marketing, 1956, 21: 3-8 CrossRef Google Scholar

[31] Azzedin F, Maheswaran M. Evolving and managing trust in grid computing systems. In: Proceedings of IEEE Canadian Conference on Electrical and Computer Engineering. Winnipeg: IEEE Press, 2002. 1424-1429. Google Scholar

[32] Travers J, Milgram S. The small world problem. Psychol Today, 1967, 1: 61--67. Google Scholar

  • Figure 1

    Purchase decision process with respect to three consumer groups.

  • Figure 2

    Four groups of consumers in online e-commerce.

  • Figure 3

    (Color online) Comparison of information gain ratio.

  • Figure 4

    The general flow of proposed method.

  • Figure 5

    The imbalanced trust relations.

  • Figure 6

    (Color online) Performance of different trust ranges. (a) MAE metric; (b) other metrics.

  • Figure 7

    (Color online) Performance of different thresholds.

  • Figure 10

    (Color online) Anomaly scores of order-consumers. (a) Among order-consumers; (b) among return-consumers.

  • Figure 11

    (Color online) Distribution of return reasons. (a) Specific return reasons; (b) unspecific return reasons.

  • Table 1   Important statistics of purchase records
    Data sources Description Statistics
    Customer Number of consumers 6223
    Product Number of products 990
    Number of orders 143835
    Order Average orders per consumer 23.11
    Average orders per product 145.29
    Number of returns or refunds 9886
    Return and refund Average returns or refunds per consumer 1.59
    Average returns or refunds per product 9.99
  • Table 2   Statistics of important attributes
    Attribute Mean Max Min Standard deviation
    Customer_credit 389.3868 25471 0 547.8237
    Active_time 1412.237 3819.027 0 749.3843
    Consumer_order_num 23.11345 78000 1 989.2386
    Consumer_return_num 1.588623 507 0 7.700736
    Item_price 81.85295 3050 0.1 113.0857
    Exist_time 580.5816 722.1351 4.003935 194.6978
    Has_warranty 0.1252458 1 0 0.3310024
    Has_invoice 0.0021157 1 0 0.0459492
    Has_showcase 0.3703141 1 0 0.482896
    Sub_stock 1.053251 2 1 0.2245371
    Has_discount 0.9733596 1 0 0.1610326
    Discount_fee 17.38692 12000 0 93.27477
    Post_fee 3.767954 5 0 2.15463
    Trade_time 6.515412 190.7472 0 4.918788
    Payment 94.29233 400000 0 3019.021
  • Table 3   Attributes description
    Feature Attribute Description
    Buyer_order Number of consumer's orders
    Consumer profile Buyer_credit Consumer's credit value
    Active_time Consumer's active time
    Exist_time Product's list time
    Product profile Item_price Product's price
    Sales_num Product's sales volume
    Payment Purchase's payment
    Transaction profile Trade_time Purchase's time
    Discount_fee Purchase's discount
  • Table 4   Symbol description
    Notation Description
    $u$,$v$ Consumer $u$,$v$ $\in$0,…,$m-1$
    $i$,$j$ Product $i$,$j$ $\in$0,…,$n-1$
    ${\rm~SimCO}$ The $n$-by-$n$ consumer order similarity graph
    ${\rm~SimCR}$ The $n$-by-$n$ consumer return similarity graph
    ${\rm~SimPR}$ The $m$-by-$m$ product return similarity graph
    $k$ The depth of the walk at the moment
    $t_0$ The initial version of consumer trust network
    $\rho$ Consumer trust network
    $\omega$ Consumer return trust relations
    $\alpha$ The stopping probability
    ${\rm~RP}_u$ The return products of the consumer $u$
    $\hat~r$ Predicted return propensity
    $r$ Return propensity
    mutual_trust The mutual-trust between two consumers
    as The anomaly score
  • Table 5   Description of purchase records
    Product ID Product price Sales volume Number of returns Number of consumers
    14064167845 88 377 53 318
  • Table 6   Description of consumer credit
    Number of returns Consumer's average credit Percent (%)
    0 400.5428 83.29
    1 316.2121 16.22
    2 116.5 0.49
    Total 385.4717 100.00
  • Table 7   Comparison of credit
    Return-consumer credit Order-consumer credit
    Selfish-potential 338.9 328
    Honest-potential 327.42 341.08
    Fraud-potential 183.83 406.7
  • Table 8   Comparison of fraud consumers
    Customer ID Trade time (day) Buyer credit Active time (day) Discount fee Return num
    l***9 6.944 81 613.53 0 3
    D***L 0.059 164 548 31 1

Copyright 2020 Science China Press Co., Ltd. 《中国科学》杂志社有限责任公司 版权所有

京ICP备18024590号-1