Beyond Movie Recommendations: Solving the Continuous Cold Start Problem in E-commerce Recommendations

J. Kiseleva, A. Tuzhilin, J. Kamps, L. Bernardi, C. Davis, I. Kovacek, M.S. Einarsen, Djoerd Hiemstra

Research output: Book/ReportReportAcademic

66 Downloads (Pure)

Abstract

Many e-commerce websites use recommender systems or personalized rankers to personalize search results based on their previous interactions. However, a large fraction of users has no prior inter-actions, making it impossible to use collaborative filtering or rely on user history for personalization. Even the most active users mayvisit only a few times a year and may have volatile needs or different personas, making their personal history a sparse and noisy signal at best. This paper investigates how, when we cannot rely on the user history, the large scale availability of other user interactions still allows us to build meaningful profiles from the contextual data and whether such contextual profiles are useful to customize the ranking, exemplified by data from a major online travel agentBooking.com.Our main findings are threefold: First, we characterize the Continuous Cold Start Problem(CoCoS) from the viewpoint of typical e-commerce applications. Second, as explicit situational con-text is not available in typical real world applications, implicit cues from transaction logs used at scale can capture essential features of situational context. Third, contextual user profiles can be created offline, resulting in a set of smaller models compared to a single huge non-contextual model, making contextual ranking available with negligible CPU and memory footprint. Finally we conclude that, in an online A/B test on live users, our contextual ranker in-creased user engagement substantially over a non-contextual base-line, with click-through-rate (CTR) increased by 20%. This clearly demonstrates the value of contextual user profiles in a real world application.
Original languageUndefined
Place of PublicationIthaca, NY, USA
PublisherCornell University
Number of pages11
DOIs
Publication statusPublished - Jul 2016

Keywords

  • EWI-27194

Cite this

Kiseleva, J., Tuzhilin, A., Kamps, J., Bernardi, L., Davis, C., Kovacek, I., ... Hiemstra, D. (2016). Beyond Movie Recommendations: Solving the Continuous Cold Start Problem in E-commerce Recommendations. Ithaca, NY, USA: Cornell University. https://doi.org/10.13140/RG.2.1.2488.7288
Kiseleva, J. ; Tuzhilin, A. ; Kamps, J. ; Bernardi, L. ; Davis, C. ; Kovacek, I. ; Einarsen, M.S. ; Hiemstra, Djoerd. / Beyond Movie Recommendations: Solving the Continuous Cold Start Problem in E-commerce Recommendations. Ithaca, NY, USA : Cornell University, 2016. 11 p.
@book{15480686994a4507a9c607e0a102c699,
title = "Beyond Movie Recommendations: Solving the Continuous Cold Start Problem in E-commerce Recommendations",
abstract = "Many e-commerce websites use recommender systems or personalized rankers to personalize search results based on their previous interactions. However, a large fraction of users has no prior inter-actions, making it impossible to use collaborative filtering or rely on user history for personalization. Even the most active users mayvisit only a few times a year and may have volatile needs or different personas, making their personal history a sparse and noisy signal at best. This paper investigates how, when we cannot rely on the user history, the large scale availability of other user interactions still allows us to build meaningful profiles from the contextual data and whether such contextual profiles are useful to customize the ranking, exemplified by data from a major online travel agentBooking.com.Our main findings are threefold: First, we characterize the Continuous Cold Start Problem(CoCoS) from the viewpoint of typical e-commerce applications. Second, as explicit situational con-text is not available in typical real world applications, implicit cues from transaction logs used at scale can capture essential features of situational context. Third, contextual user profiles can be created offline, resulting in a set of smaller models compared to a single huge non-contextual model, making contextual ranking available with negligible CPU and memory footprint. Finally we conclude that, in an online A/B test on live users, our contextual ranker in-creased user engagement substantially over a non-contextual base-line, with click-through-rate (CTR) increased by 20{\%}. This clearly demonstrates the value of contextual user profiles in a real world application.",
keywords = "EWI-27194",
author = "J. Kiseleva and A. Tuzhilin and J. Kamps and L. Bernardi and C. Davis and I. Kovacek and M.S. Einarsen and Djoerd Hiemstra",
year = "2016",
month = "7",
doi = "10.13140/RG.2.1.2488.7288",
language = "Undefined",
publisher = "Cornell University",
address = "United States",

}

Kiseleva, J, Tuzhilin, A, Kamps, J, Bernardi, L, Davis, C, Kovacek, I, Einarsen, MS & Hiemstra, D 2016, Beyond Movie Recommendations: Solving the Continuous Cold Start Problem in E-commerce Recommendations. Cornell University, Ithaca, NY, USA. https://doi.org/10.13140/RG.2.1.2488.7288

Beyond Movie Recommendations: Solving the Continuous Cold Start Problem in E-commerce Recommendations. / Kiseleva, J.; Tuzhilin, A.; Kamps, J.; Bernardi, L.; Davis, C.; Kovacek, I.; Einarsen, M.S.; Hiemstra, Djoerd.

Ithaca, NY, USA : Cornell University, 2016. 11 p.

Research output: Book/ReportReportAcademic

TY - BOOK

T1 - Beyond Movie Recommendations: Solving the Continuous Cold Start Problem in E-commerce Recommendations

AU - Kiseleva, J.

AU - Tuzhilin, A.

AU - Kamps, J.

AU - Bernardi, L.

AU - Davis, C.

AU - Kovacek, I.

AU - Einarsen, M.S.

AU - Hiemstra, Djoerd

PY - 2016/7

Y1 - 2016/7

N2 - Many e-commerce websites use recommender systems or personalized rankers to personalize search results based on their previous interactions. However, a large fraction of users has no prior inter-actions, making it impossible to use collaborative filtering or rely on user history for personalization. Even the most active users mayvisit only a few times a year and may have volatile needs or different personas, making their personal history a sparse and noisy signal at best. This paper investigates how, when we cannot rely on the user history, the large scale availability of other user interactions still allows us to build meaningful profiles from the contextual data and whether such contextual profiles are useful to customize the ranking, exemplified by data from a major online travel agentBooking.com.Our main findings are threefold: First, we characterize the Continuous Cold Start Problem(CoCoS) from the viewpoint of typical e-commerce applications. Second, as explicit situational con-text is not available in typical real world applications, implicit cues from transaction logs used at scale can capture essential features of situational context. Third, contextual user profiles can be created offline, resulting in a set of smaller models compared to a single huge non-contextual model, making contextual ranking available with negligible CPU and memory footprint. Finally we conclude that, in an online A/B test on live users, our contextual ranker in-creased user engagement substantially over a non-contextual base-line, with click-through-rate (CTR) increased by 20%. This clearly demonstrates the value of contextual user profiles in a real world application.

AB - Many e-commerce websites use recommender systems or personalized rankers to personalize search results based on their previous interactions. However, a large fraction of users has no prior inter-actions, making it impossible to use collaborative filtering or rely on user history for personalization. Even the most active users mayvisit only a few times a year and may have volatile needs or different personas, making their personal history a sparse and noisy signal at best. This paper investigates how, when we cannot rely on the user history, the large scale availability of other user interactions still allows us to build meaningful profiles from the contextual data and whether such contextual profiles are useful to customize the ranking, exemplified by data from a major online travel agentBooking.com.Our main findings are threefold: First, we characterize the Continuous Cold Start Problem(CoCoS) from the viewpoint of typical e-commerce applications. Second, as explicit situational con-text is not available in typical real world applications, implicit cues from transaction logs used at scale can capture essential features of situational context. Third, contextual user profiles can be created offline, resulting in a set of smaller models compared to a single huge non-contextual model, making contextual ranking available with negligible CPU and memory footprint. Finally we conclude that, in an online A/B test on live users, our contextual ranker in-creased user engagement substantially over a non-contextual base-line, with click-through-rate (CTR) increased by 20%. This clearly demonstrates the value of contextual user profiles in a real world application.

KW - EWI-27194

U2 - 10.13140/RG.2.1.2488.7288

DO - 10.13140/RG.2.1.2488.7288

M3 - Report

BT - Beyond Movie Recommendations: Solving the Continuous Cold Start Problem in E-commerce Recommendations

PB - Cornell University

CY - Ithaca, NY, USA

ER -

Kiseleva J, Tuzhilin A, Kamps J, Bernardi L, Davis C, Kovacek I et al. Beyond Movie Recommendations: Solving the Continuous Cold Start Problem in E-commerce Recommendations. Ithaca, NY, USA: Cornell University, 2016. 11 p. https://doi.org/10.13140/RG.2.1.2488.7288