Factored MDPs for detecting topics of user sessions

Research output: Contributions to collected editions/worksArticle in conference proceedingsResearchpeer-review

Standard

Factored MDPs for detecting topics of user sessions. / Tavakol, Maryam; Brefeld, Ulf.
RecSys 2014 - Proceedings of the 8th ACM Conference on Recommender Systems. Association for Computing Machinery, Inc, 2014. p. 33-40.

Research output: Contributions to collected editions/worksArticle in conference proceedingsResearchpeer-review

Harvard

Tavakol, M & Brefeld, U 2014, Factored MDPs for detecting topics of user sessions. in RecSys 2014 - Proceedings of the 8th ACM Conference on Recommender Systems. Association for Computing Machinery, Inc, pp. 33-40, 8th ACM Conference on Recommender Systems - RecSys2014, Foster City, United States, 06.10.14. https://doi.org/10.1145/2645710.2645739

APA

Tavakol, M., & Brefeld, U. (2014). Factored MDPs for detecting topics of user sessions. In RecSys 2014 - Proceedings of the 8th ACM Conference on Recommender Systems (pp. 33-40). Association for Computing Machinery, Inc. https://doi.org/10.1145/2645710.2645739

Vancouver

Tavakol M, Brefeld U. Factored MDPs for detecting topics of user sessions. In RecSys 2014 - Proceedings of the 8th ACM Conference on Recommender Systems. Association for Computing Machinery, Inc. 2014. p. 33-40 doi: 10.1145/2645710.2645739

Bibtex

@inbook{9b7a16a7bd684cb6bb1dd193ced6e3e7,
title = "Factored MDPs for detecting topics of user sessions",
abstract = "Recommender systems aim to capture interests of users to provide tailored recommendations. User interests are however often unique and depend on many unobservable factors including a user's mood and the local weather. We take a contextual session-based approach and propose a sequential framework using factored Markov decision processes (fMDPs) to detect the user's goal (the topic) of a session. We show that an independence assumption on the attributes of items leads to a set of independent models that can be optimised efficiently. Our approach results in interpretable topics that can be effectively turned into recommendations. Empirical results on a real world click log from a large e-commerce company exhibit highly accurate topic prediction rates of about 90%. Translating our approach into a topic-driven recommender system outperforms several baseline competitors.",
keywords = "Informatics, MDP, Recommender systems, Session-based, User intent, Business informatics",
author = "Maryam Tavakol and Ulf Brefeld",
year = "2014",
month = oct,
day = "6",
doi = "10.1145/2645710.2645739",
language = "English",
isbn = "978-1-4503-2668-1",
pages = "33--40",
booktitle = "RecSys 2014 - Proceedings of the 8th ACM Conference on Recommender Systems",
publisher = "Association for Computing Machinery, Inc",
address = "United States",
note = "8th ACM Conference on Recommender Systems - RecSys2014, RecSys 2014 ; Conference date: 06-10-2014 Through 10-10-2014",
url = "https://recsys.acm.org/recsys14/",

}

RIS

TY - CHAP

T1 - Factored MDPs for detecting topics of user sessions

AU - Tavakol, Maryam

AU - Brefeld, Ulf

N1 - Conference code: 8

PY - 2014/10/6

Y1 - 2014/10/6

N2 - Recommender systems aim to capture interests of users to provide tailored recommendations. User interests are however often unique and depend on many unobservable factors including a user's mood and the local weather. We take a contextual session-based approach and propose a sequential framework using factored Markov decision processes (fMDPs) to detect the user's goal (the topic) of a session. We show that an independence assumption on the attributes of items leads to a set of independent models that can be optimised efficiently. Our approach results in interpretable topics that can be effectively turned into recommendations. Empirical results on a real world click log from a large e-commerce company exhibit highly accurate topic prediction rates of about 90%. Translating our approach into a topic-driven recommender system outperforms several baseline competitors.

AB - Recommender systems aim to capture interests of users to provide tailored recommendations. User interests are however often unique and depend on many unobservable factors including a user's mood and the local weather. We take a contextual session-based approach and propose a sequential framework using factored Markov decision processes (fMDPs) to detect the user's goal (the topic) of a session. We show that an independence assumption on the attributes of items leads to a set of independent models that can be optimised efficiently. Our approach results in interpretable topics that can be effectively turned into recommendations. Empirical results on a real world click log from a large e-commerce company exhibit highly accurate topic prediction rates of about 90%. Translating our approach into a topic-driven recommender system outperforms several baseline competitors.

KW - Informatics

KW - MDP

KW - Recommender systems

KW - Session-based

KW - User intent

KW - Business informatics

UR - http://www.scopus.com/inward/record.url?scp=84908893447&partnerID=8YFLogxK

U2 - 10.1145/2645710.2645739

DO - 10.1145/2645710.2645739

M3 - Article in conference proceedings

AN - SCOPUS:84908893447

SN - 978-1-4503-2668-1

SP - 33

EP - 40

BT - RecSys 2014 - Proceedings of the 8th ACM Conference on Recommender Systems

PB - Association for Computing Machinery, Inc

T2 - 8th ACM Conference on Recommender Systems - RecSys2014

Y2 - 6 October 2014 through 10 October 2014

ER -

DOI