The dynamics of repeat consumption

Abstract

We study the patterns by which a user consumes the same item repeatedly over time, in a wide variety domains ranging from checkins at the same business location to re-watches of the same video. We find that recency of consumption is the strongest predictor of repeat consumption. Based on this, we develop a model by which the item from t timesteps ago is reconsumed with a probability proportional to a function of t. We study theoretical properties of this model, develop algorithms to learn reconsumption likelihood as a function of t, and show a strong fit of the resulting inferred function via a power law with exponential cutoff. We then introduce a notion of item quality, show that it alone underperforms our recency-based model, and develop a hybrid model that predicts user choice based on a combination of recency and quality. We show how the parameters of this model may be jointly estimated, and show that the resulting scheme outperforms other alternatives.

On the Dynamics of Social Media Popularity

“…The existence of Cluster 0 can be attributed to three possible reasons. First, there are certain topics that users will continue to revisit over time [Anderson et al 2014;Wang and Huberman 2012], and thus the content will not follow a rise-and-fall pattern (as proposed in [Matsubara et al 2012]). Second, the propagation of these topics is much slower [Wang and Huberman 2012], being the pattern we see still part of the growth period in interest in that particular topic.…”

“…Last, YouTube's own growth in popularity over time may cause the audience of interest in some videos to increase. Intuitively, a combination of these factors will likely be the case, and only recently researchers have started looking into the implications of each of them [Anderson et al 2014;Wang and Huberman 2012].…”

Repeat Consumption Recommendation Based on Users Preference Dynamics and Side Information

“…Future work includes the incorporation of items' inherent quality of attractiveness into our models, which may increase the recommendation accuracy according to the analysis of the dynamics of repeat consumption by Anderson et al in [2].…”

“…Temporal collaborative filtering with matrix [4,7] and tensor factorization techniques [3] can generate accurate recommendation, since they can capture the drifts in the rating behavior and the changes of user preference over time in a collaborative-filtering fashion. However, users may repeatedly interact with items over time in several applications [2]; for instance, visiting the same web sites, buying retail items from Amazon or implicitly interacting, such as artist listenings on last.fm or movie viewing from a specific genre on MovieLens. To account also for the fact that users' side information, such as demographics, can improve the recommendation accuracy [5], we present a basic CTF model and its variant W-CTF, where the diversity of Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage, and that copies bear this notice and the full citation on the first page.…”

Modeling the Interest-Forgetting Curve for Music Recommendation

“…Besides, recency effects have also been observed in people's listening to music [1,7], where people have a tendency to recall the recently heard songs other than those heard long ago. A straightforward intuition is that songs consumed more recently contribute more to the understanding of current user context, which leads to non-increasing weights as the consumption timestamps get farther from now.…”