Recurrent Recommendation with Local Coherence Jianling Wang and - PowerPoint PPT Presentation

Recurrent Recommendation with Local Coherence Jianling Wang and James Caverlee

Dynamics in Recommenders • Users and Items are constantly in flux. … …

Local Coherence • Within a short-term sequence, the neighboring items or users is likely to be coherent . … … …

Local Coherence • Within a short-term sequence, the neighboring items or users is likely to be coherent . … … … Coherent Items:

Problem • Balance the local coherence with long-term evolution among both implicit and explicit feedback.

Our Goal • Predict how users will rate items in the future . ?

Our Approach R ecurrent R ecommendation with L ocal C oherence - RRLC

Challenge 1: Feedback Sparsity • We want to model the consistent latent factors of users and items with their interactions. Consistent Consistent Embedding Embedding

Challenge 1: Feedback Sparsity • The interactions between users and items are sparse . Sparse Data ! Consistent Consistent Embedding Embedding

Solution • Coherence-based Neighbors. Sparse Data ! Coherence-based Coherence-based Neighboring Neighboring Users Items Consistent Consistent { { Embedding Embedding

Challenge 2: Representations • We use Recurrent Neural Networks (RNN) to model the dynamics of users and items. Dynamic Dynamic … … ∆ T ∆ T Item-RNN User-RNN

Challenge 2: Representations • We want e ff ective representations of rating events . Dynamic Dynamic ? … … ? ∆ T ∆ T Item-RNN User-RNN

Solution • Rating Event Embedding utilizing local coherence Dynamic Dynamic Coherence-based Coherence-based Rating Event Rating Event … … Embedding Embedding ∆ T ∆ T Item-RNN User-RNN

RRLC vs. Neural Models • ~6% Improvement Time-independent Neural Models:

RRLC vs. Time-Dependent Models • ~5% Improvement Time-dependent Models:

Conclusion • Propose a novel RNN-based recommender • Enhanced with local coherence on both implicit and explicit feedback sequences • Outperform state-of-art models in ratings prediction • Help to alleviate the cold start problem Thank you! Jianling Wang and James Caverlee Texas A&M University, USA