-
Semi-supervised Collaborative Ranking with Push at Top
Authors:
Iman Barjasteh,
Rana Forsati,
Abdol-Hossein Esfahanian,
Hayder Radha
Abstract:
Existing collaborative ranking based recommender systems tend to perform best when there is enough observed ratings for each user and the observation is made completely at random. Under this setting recommender systems can properly suggest a list of recommendations according to the user interests. However, when the observed ratings are extremely sparse (e.g. in the case of cold-start users where n…
▽ More
Existing collaborative ranking based recommender systems tend to perform best when there is enough observed ratings for each user and the observation is made completely at random. Under this setting recommender systems can properly suggest a list of recommendations according to the user interests. However, when the observed ratings are extremely sparse (e.g. in the case of cold-start users where no rating data is available), and are not sampled uniformly at random, existing ranking methods fail to effectively leverage side information to transduct the knowledge from existing ratings to unobserved ones. We propose a semi-supervised collaborative ranking model, dubbed \texttt{S$^2$COR}, to improve the quality of cold-start recommendation. \texttt{S$^2$COR} mitigates the sparsity issue by leveraging side information about both observed and missing ratings by collaboratively learning the ranking model. This enables it to deal with the case of missing data not at random, but to also effectively incorporate the available side information in transduction. We experimentally evaluated our proposed algorithm on a number of challenging real-world datasets and compared against state-of-the-art models for cold-start recommendation. We report significantly higher quality recommendations with our algorithm compared to the state-of-the-art.
△ Less
Submitted 16 November, 2015;
originally announced November 2015.
-
The Entropy of Attention and Popularity in YouTube Videos
Authors:
Jonathan Scott Morgan,
Iman Barjasteh,
Cliff Lampe,
Hayder Radha
Abstract:
The vast majority of YouTube videos never become popular, languishing in obscurity with few views, no likes, and no comments. We use information theoretical measures based on entropy to examine how time series distributions of common measures of popularity in videos from YouTube's "Trending videos" and "Most recent" video feeds relate to the theoretical concept of attention. While most of the vide…
▽ More
The vast majority of YouTube videos never become popular, languishing in obscurity with few views, no likes, and no comments. We use information theoretical measures based on entropy to examine how time series distributions of common measures of popularity in videos from YouTube's "Trending videos" and "Most recent" video feeds relate to the theoretical concept of attention. While most of the videos in the "Most recent" feed are never popular, some 20% of them have distributions of attention metrics and measures of entropy that are similar to distributions for "Trending videos". We analyze how the 20% of "Most recent" videos that become somewhat popular differ from the 80% that do not, then compare these popular "Most recent" videos to different subsets of "Trending videos" to try to characterize and compare the attention each receives.
△ Less
Submitted 2 December, 2014;
originally announced December 2014.
-
Trending Videos: Measurement and Analysis
Authors:
Iman Barjasteh,
Ying Liu,
Hayder Radha
Abstract:
Unlike popular videos, which would have already achieved high viewership numbers by the time they are declared popular, YouTube trending videos represent content that targets viewers attention over a relatively short time, and has the potential of becoming popular. Despite their importance and visibility, YouTube trending videos have not been studied or analyzed thoroughly. In this paper, we prese…
▽ More
Unlike popular videos, which would have already achieved high viewership numbers by the time they are declared popular, YouTube trending videos represent content that targets viewers attention over a relatively short time, and has the potential of becoming popular. Despite their importance and visibility, YouTube trending videos have not been studied or analyzed thoroughly. In this paper, we present our findings for measuring, analyzing, and comparing key aspects of YouTube trending videos. Our study is based on collecting and monitoring high-resolution time-series of the viewership and related statistics of more than 8,000 YouTube videos over an aggregate period of nine months. Since trending videos are declared as such just several hours after they are uploaded, we are able to analyze trending videos time-series across critical and sufficiently-long durations of their lifecycle. In addition, we analyze the profile of users who upload trending videos, to potentially identify the role that these users profile plays in getting their uploaded videos trending. Furthermore, we conduct a directional-relationship analysis among all pairs of trending videos time-series that we have monitored. We employ Granger Causality (GC) with significance testing to conduct this analysis. Unlike traditional correlation measures, our directional-relationship analysis provides a deeper insight onto the viewership pattern of different categories of trending videos. Trending videos and their channels have clear distinct statistical attributes when compared to other YouTube content that has not been labeled as trending. Our results also reveal a highly asymmetric directional-relationship among different categories of trending videos. Our directionality analysis also shows a clear pattern of viewership toward popular categories, whereas some categories tend to be isolated.
△ Less
Submitted 26 September, 2014;
originally announced September 2014.
