Fusing Similarity Models with Markov Chains for Sparse Sequential Recommendation

By Ruining He, Julian McAuley. Industrial Conference on Data Mining, 2016

882 citations recorded by the source corpus.

Predicting personalized sequential behavior is a key task for recommender systems. In order to predict user actions such as the next product to purchase, movie to watch, or place to visit, it is essential to take into account both long-term user preferences and sequential patterns (i.e., short-term dynamics). Matrix Factorization and Markov Chain methods have emerged as two separate but powerful paradigms for modeling the two respectively. Combining these ideas has led to unified methods that accommodate long-and short-term dynamics simultaneously by modeling pairwise user-item and item-item interactions. In spite of the success of such methods for tackling dense data, they are challenged by sparsity issues, which are prevalent in real-world datasets. In recent years, similarity-based methods have been proposed for (sequentially-unaware) item recommendation with promising results on sparse datasets. In this paper, we propose to fuse such methods with Markov Chains to make personalized sequential recommendations. We evaluate our method, Fossil, on a variety of large, real-world datasets. We show quantitatively that Fossil outperforms alternative algorithms, especially on sparse datasets, and qualitatively that it captures personalized dynamics and is able to make meaningful recommendations.

Subfields: Artificial Intelligence, Information Systems, Management Science and Operations Research

View Paper: https://www.semanticscholar.org/paper/fea7e2bcbe852bcb33ce70a4d57664e954f0e82a

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Xiangnan He
2. Wayne Xin Zhao
3. Yongfeng Zhang
4. Ji-rong Wen
5. Xu Chen
6. Julian McAuley
7. Yupeng Hou
8. Xiang Wang

In Information Systems

1. Xiangnan He
2. Yongfeng Zhang
3. Wayne Xin Zhao
4. Xu Chen
5. Ji-rong Wen
6. Xiang Wang
7. Yupeng Hou
8. Julian McAuley

In Management Science and Operations Research

1. Dietmar Jannach
2. Malte Ludewig
3. Jiaxi Tang
4. Ke Wang
5. Xiangnan He
6. Yan Wang
7. Longbing Cao
8. Quan Z. Sheng

Citations in the corpus

• Contrastive Learning for Sequential Recommendation (2022, neutral): “To consider more preceding items and extract more complicated patterns, the high-order MC was also explored in the following work [16].”
• RecBole: Towards a Unified, Comprehensive and Efficient Framework for Recommendation Algorithms (2020, neutral): “Fossil [18] ICDM 2016 Amazon, Epinions, Foursquare”
• Sequential Recommender System based on Hierarchical Attention Networks (2018, methodology): “In the literature, researchers usually employ separate models to characterize user’s long-term preference (i.e, general taste) and short-term preference (i.e., sequential pattern), and then integrate them together [Rendle et al. , 2009; Feng et al. , 2015; He and McAuley, 2016].”
• STAN: Spatio-Temporal Attention Network for Next Location Recommendation (2021, methodology): “Matrix factorization models [15, 26] are proposed to approach this problem, with further extensions [3, 10] find that explicit spatial and temporal information help a lot with recommendation performance.”
• Sequential Recommendation with User Memory Networks (2018, background): “Tomodelmultiple-step sequential behaviors, [10] adoptedMarkov chains to provide recommendations with sparse sequences, and [31] proposed the dynamic recurrent basket model (DREAM) to capture global sequential patterns and to learn dynamic user interest representations based on recurrent neural netwo”
• Long and Short-Term Recommendations with Recurrent Neural Networks (2017, methodology): “Recently however, several papers have explored sequencebased recommendation, mostly motivated by the need to produce recommendations without building a user model[8, 9, 20].”
• BERT4Rec: Sequential Recommendation with Bidirectional Encoder Representations from Transformer (2019, methodology): “Besides the first-order MCs, high-order MCs are also adopted to consider more previous items [10, 11].”
• A Survey on Session-based Recommender Systems (2019, background): “FOSSIL [39] Next item Latent factor ICDM 2016 https://github.”
• TALLRec: An Effective and Efficient Tuning Framework to Align Large Language Model with Recommendation (2023, background): “In the early time, the Markov chain plays an important role in sequential recommendation [13, 31, 39, 48].”
• Adapting Large Language Models by Integrating Collaborative Semantics for Recommendation (2023, methodology): “Recently, typical methods become to adopt various deep neural networks to model user historical behaviors represented by item ID sequences, including RNN [1], [3], CNN [4], [5], GNN [6], [7], and Transformer [2], [8].”
• Intent Contrastive Learning for Sequential Recommendation (2022, background): “FPMC [34] combines the advantages of Markov Chains and matrix factorization to fuse both sequential patterns and users’ general interest.”
• Personalized Top-N Sequential Recommendation via Convolutional Sequence Embedding (2018, methodology): “Although Fossil [6] considers a high-order Markov chain, the overall in uence is a weighted sum of previous items’ latent representations factorized from rstorder Markov transition matrices.”
• Graph Neural Networks in Recommender Systems: A Survey (2020, neutral): “Some works adopt Markov Chain (MC) [18, 44] to capture the item-to-item transition based on the assumption that the most recent clicked item reflects the user’s dynamic preference.”
• Time Interval Aware Self-Attention for Sequential Recommendation (2020, methodology): “PRME [6], HRM [27], Fossil [8], GRU4Rec[12]) have been outperformed by the baselines above, so we omit comparison against them.”
• Evaluation of session-based recommendation algorithms (2018, methodology): “Later on, variations of fpmc were proposed and successfully applied for a variety of application problems, e.g., by [Kabbur et al. 2013] and [He and McAuley 2016].”
• Hierarchical Gating Networks for Sequential Recommendation (2019, methodology): “Given our extensive comparisons against the state-of-the-art methods, we omit comparisons with methods such as FMC and FPMC [30], Fossil [8], since they have been outperformed by the recently proposed Caser and SASRec.”
• Adversarial and Contrastive Variational Autoencoder for Sequential Recommendation (2021, methodology): “Following FPMC, higher-order Markov chain [3] is used in learning high-order sequential dependencies.”
• Controllable Multi-Interest Framework for Recommendation (2020, methodology): “Fossil [17] integrates similarity-based methods with Markov Chains smoothly to make personalized sequential predictions on sparse and long-tailed datasets.”