Abstract
This paper adopts a Bayesian approach for finding top recommendations. The approach is entirely personalized, and consists of learning a utility function over user preferences via employing a sampling-based, non-intrusive preference elicitation framework. We explicitly model the uncertainty over the utility function and learn it through passive user feedback, provided in the form of clicks on previously recommended items. The utility function is a linear combination of weighted features, and beliefs are maintained using a Markov Chain Monte Carlo algorithm. Our approach overcomes the problem of having conflicting user constraints by identifying a convex region within a user’s preferences model. Additionally, it handles situations where not enough data about the user is available, by exploiting the information from clusters of (feature) weight vectors created by observing other users’ behavior. We evaluate our system’s performance by applying it in the online hotel booking recommendations domain using a real-world dataset, with very encouraging results.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
- 2.
To illustrate, assume a “hotel price” is “low”, say 0.1. If the user prefers really cheap hotels, she might have a weight of \(-0.9\) for “hotel price”, thus deriving a higher utility for this hotel compared to that derived for an expensive hotel of price, say, 0.9 (since \(-0.09 > -0.81\)).
- 3.
Moreover, all clicked items are originally equally appealing. However, as interactions with the system increase, beliefs about the desirability of the items get updated.
- 4.
We note that ours is essentially a standard version of the Metropolis-Hastings algorithm, which is known to be almost always convergent [21].
- 5.
In some detail, we divide each of the m dimensions into a fixed number of “segments”–ten (10) in our implementation—and use this segmentation to generate “buckets” to place our samples into. In this way, we create \(10^m\) buckets in total: for instance, if we had only two dimensions, e.g. “price” and “distance to city center”, we would be creating 100 buckets. Then, each prior sample is allocated to its corresponding bucket, based on Euclidean distance. When Algorithm 1 picks a sample, it checks which bucket it belongs to, and uses the number of samples in the buckets to estimate the \(\pi (\cdot )\) density in Eq. 7. Thus, this method uses the prior samples to estimate the posterior joint density.
- 6.
Specifically, 2 out of 7 items presented to the user are chosen randomly; see Sect. 4.2 below.
- 7.
The idea of employing clustering to address the “cold start” problem has also appeared in [26]. However, that work uses averaging over user ratings to produce recommendations that are appropriate for each cluster. In our work, we make no use of user ratings over items, and make recommendations based on the clusters’ centroids rather than employing some averaging-over-cluster-contents process.
- 8.
Note that “clients’ rating” is just an item’s (a hotel’s) feature. We stress that we do not ask our system’s users to rate the items, and it is not the system’s aim to produce recommendations based on such ratings.
- 9.
References
Andrieu, C., de Freitas, N., Doucet, A., Jordan, M.I.: An introduction to MCMC for machine learning. Mach. Learn. 50(1), 5–43 (2003). https://doi.org/10.1023/A:1020281327116
Applegate, D., Kannan, R.: Sampling and integration of near log-concave functions. In: Proceedings of the Twenty-Third Annual ACM Symposium on Theory of Computing, pp. 156–163 (1991)
Babas, K., Chalkiadakis, G., Tripolitakis, E.: You are what you consume: a Bayesian method for personalized recommendations. In: Proceedings of the 7th ACM Conference on Recommender Systems (ACM RecSys 2013), Hong Kong, China (2013)
Bishop, C.M.: Neural networks for pattern recognition (1995)
Boutilier, C.: A POMDP formulation of preference elicitation problems. In: Proceedings of the 18th AAAI Conference, AAAI 2002 (2002)
Bowling, M.H., Veloso, M.M.: Multiagent learning using a variable learning rate. Artif. Intell. (AIJ) 136(2), 215–250 (2002)
Chajewska, U., Koller, D., Parr, R.: Making rational decisions using adaptive utility elicitation. In: Proceedings of the 17th AAAI Conference, AAAI 2000, pp. 363–369 (2000)
Chajewska, U., Koller, D., Ormoneit, D.: Learning an agent’s utility function by observing behavior. In: Proceedings of the International Conference on Machine Learning (ICML) (2001)
Chib, S., Greenberg, E.: Understanding the Metropolis-Hastings algorithm, pp. 327–335 (2012)
Dantzig, G.B., Orden, A., Wolfe, P.: The generalized simplex method for minimizing a linear form under linear inequality restraints. Pacific J. Math. 5(2), 183–195 (1955)
DeGroot, M.: Probability and Statistics. Addison-Wesley Series in Behavioral Science, Addison-Wesley Publishing Company, Boston (1975). https://books.google.gr/books?id=fxPvAAAAMAAJ
Dong, R., Smyth, B.: From more-like-this to better-than-this: hotel recommendations from user generated reviews. In: Proceedings of the 2016 Conference on User Modeling Adaptation and Personalization (UMAP 2016), pp. 309–310 (2016)
Gilks, W.R., Wild, P.: Adaptive rejection sampling for Gibbs sampling. J. Roy. Stat. Soc. Ser. C (Appl. Stat.) 41(2), 337–348 (1992)
Hastings, W.K.: Monte Carlo sampling methods using Markov chains and their applications. Biometrika 57(1), 97–109 (1970). http://www.jstor.org/stable/2334940
Keeney, R.L., Raiffa, H.: Decisions with Multiple Objectives: Decisions with Preferences and Value Tradeoffs. Cambridge University Press, Cambridge (1993)
MacQueen, J.B.: Some methods for classification and analysis of multivariate observations. In: Proceedings of 5th Berkeley Symposium on Mathematical Statistics and Probability (1967)
Nasery, M., Braunhofer, M., Ricci, F.: Recommendations with optimal combination of feature-based and item-based preferences. In: Proceedings of the 2016 Conference on User Modeling Adaptation and Personalization (UMAP 2016), pp. 269–273 (2016)
Ng, A.Y., Russell, S.J.: Algorithms for inverse reinforcement learning. In: Proceedings of the Seventeenth International Conference on Machine Learning, pp. 663–670. ICML (2000)
Ramachandran, D., Amir, E.: Bayesian inverse reinforcement learning. In: Proceedings of IJCAI-2007, pp. 2586–2591 (2007)
Ricci, F., Rokach, L., Shapira, B., Kantor, P.B.: Recommender Systems Handbook, 1st edn. Springer, New York (2010). https://doi.org/10.1007/978-0-387-85820-3
Robert, C.P., Casella, G.: Monte Carlo Statistical Methods. Springer, Heidelberg (2004). https://doi.org/10.1007/978-1-4757-4145-2
Roy, S.B., Das, G., Amer-Yahia, S., Yu, C.: Interactive itinerary planning. In: Proceedings of the IEEE International Conference on Data Engineering (ICDE) (2011)
Russell, S., Norvig, P.: Artificial Intelligence: A Modern Approach, 3rd edn. Prentice Hall, Upper Saddle River (2009)
Tripolitakis, E., Chalkiadakis, G.: Probabilistic topic modeling, reinforcement learning, and crowdsourcing for personalized recommendations. In: Criado Pacheco, N., Carrascosa, C., Osman, N., Julián Inglada, V. (eds.) EUMAS/AT -2016. LNCS (LNAI), vol. 10207, pp. 157–171. Springer, Cham (2017). https://doi.org/10.1007/978-3-319-59294-7_14
Xie, M., Lakshmanan, L.V., Wood, P.T.: Generating top-k packages via preference elicitation. Proc. VLDB Endow. 7(14), 1941–1952 (2014)
Yanxiang, L., Deke, G., Fei, C., Honghui, C.: User-based clustering with top-n recommendation on cold-start problem. In: 2013 Third International Conference on Intelligent System Design and Engineering Applications (2013)
Acknowledgements
The authors would like to thank Professor Michail Lagoudakis for extremely useful suggestions for improving an earlier version of this work.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2019 Springer Nature Switzerland AG
About this paper
Cite this paper
Papilaris, MA., Chalkiadakis, G. (2019). Markov Chain Monte Carlo for Effective Personalized Recommendations. In: Slavkovik, M. (eds) Multi-Agent Systems. EUMAS 2018. Lecture Notes in Computer Science(), vol 11450. Springer, Cham. https://doi.org/10.1007/978-3-030-14174-5_13
Download citation
DOI: https://doi.org/10.1007/978-3-030-14174-5_13
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-14173-8
Online ISBN: 978-3-030-14174-5
eBook Packages: Computer ScienceComputer Science (R0)