When Recurrent Neural Networks meet the Neighborhood for Session-Based Recommendation

By Dietmar Jannach, Malte Ludewig. ACM Conference on Recommender Systems, 2017

494 citations recorded by the source corpus.

Deep learning methods have led to substantial progress in various application fields of AI, and in recent years a number of proposals were made to improve recommender systems with artificial neural networks. For the problem of making session-based recommendations, i.e., for recommending the next item in an anonymous session, Hidasi et al.~recently investigated the application of recurrent neural networks with Gated Recurrent Units (GRU4REC). Assessing the true effectiveness of such novel approaches based only on what is reported in the literature is however difficult when no standard evaluation protocols are applied and when the strength of the baselines used in the performance comparison is not clear. In this work we show based on a comprehensive empirical evaluation that a heuristics-based nearest neighbor (kNN) scheme for sessions outperforms GRU4REC in the large majority of the tested configurations and datasets. Neighborhood sampling and efficient in-memory data structures ensure the scalability of the kNN method. The best results in the end were often achieved when we combine the kNN approach with GRU4REC, which shows that RNNs can leverage sequential signals in the data that cannot be detected by the co-occurrence-based kNN method.

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

View Paper: https://www.semanticscholar.org/paper/4ab5a2b7eab033c2c2059b6e1e3e6f3e89ebd95f

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

In Information Systems

In Management Science and Operations Research

Citations in the corpus

Recommendation system based on deep learning methods: a systematic review and new directions (2019, background): “Hidasi et al. (2016b), Bansal et al. (2016), Jannach and Ludewig (2017) and Ludewig and Jannach (2018) exploited RNN architecture to better extract features for the user’s immediate next actions recommendation.”
Sequence-Aware Recommender Systems (2018, background): “ns where item recency plays a role, including news, music and e-commerce, being able to detect and leverage short-term trends in a community is highly important. For the e-commerce domain, works like [52] or [62] show that focusing on items that were trending on the platform on the last few days can”
A Survey on Accuracy-Oriented Neural Recommendation: From Collaborative Filtering to Information-Rich Recommendation (2021, methodology): “…[212] Sequential Models RNN based models that rely on sessions to construct input and output p-RNN [202],KERL [193],CRNNs [213], NARM [214], [215], [216], [217], [201], Translation based models for modeling the correlations of consecutive items TransRec [46],PeterRec [218], [219] Convolutional…”
Evaluation of session-based recommendation algorithms (2018, methodology): “Except for the RSC15-S dataset, which we include to make our evaluation comparable with previous works [Hidasi and Karatzoglou 2017; Jannach and Ludewig 2017], we report the average values after creating five data splits as described above.”
News recommender systems - Survey and roads ahead (2018, background): “, in [81, 82]), remains a challenge, in particular when the models are based on deep learning strategies.”
Personalized Top-N Sequential Recommendation via Convolutional Sequence Embedding (2018, methodology): “Recurrent neural networks (RNN) was used for session-based recommendation [8, 10].”
ADER: Adaptively Distilled Exemplar Replay Towards Continual Learning for Session-based Recommendation (2020, result): “[15] proposes SKNN to compare the current session with historical sessions in the training data.”
Hierarchical Temporal Convolutional Networks for Dynamic Recommender Systems (2019, methodology): “architecture for dynamic recommender systems [6, 12, 16, 23, 34].”
Are we really making much progress? A worrying analysis of recent neural recommendation approaches (2019, background): “However, with appropriate data pre-processing and data samplingmechanisms, scalability can also be ensured for such methods, both in academic and industrial environments [19, 26].”
Deep Learning Based Recommender System (2017, result): “RNNs [5, 28, 35, 56, 57, 73, 78, 90, 117, 132, 139, 142, 174–176], [24, 29, 33, 55, 68, 91, 108, 113, 133, 141, 149, 173, 179]”
Rethinking the Item Order in Session-based Recommendation with Graph Neural Networks (2019, methodology): “In addition to RNN, Jannach and Ludewig [11] proposed to use the neighborhood-based method to capture cooccurrence signals.”
Session-based Recommendation with Graph Neural Networks (2018, methodology): “Jannach and Ludewig (2017) combine the recurrent method and the neighborhood-based method together to mix the sequential patterns and co-occurrence signals.”
Self-Supervised Hypergraph Convolutional Networks for Session-based Recommendation (2020, background): “…the research efforts in this area regard the sessions as ordered sequences, among which recurrent neural networks (RNNs) based (Hidasi et al. 2015; Jannach and Ludewig 2017; Hidasi and Karatzoglou 2018) and graph neural networks (GNNs) (Wu et al. 2020) based approaches have shown great…”
Global Context Enhanced Graph Neural Networks for Session-based Recommendation (2020, methodology): “KNNRNN [3] makes use of GRU4REC [2] and the co-occurrence-based KNN model to extract the sequential patterns for SBR. Recently, Wang et al. [19] propose an end-to-end neural model named CSRM, which achieves state-of-the-art performance.”
A Survey on Session-based Recommender Systems (2019, methodology): “…[6, 44]; (3) time decay or attention mechanism into RNN to discriminate the intra-session dependencies and thus achieve more precise recommendations [7, 85]; and (4) traditional models like factorization machines or neighbourhood models to make up the drawbacks of RNN-only models [51, 114].”