Temporal Linear Item-Item Model for Sequential Recommendation

TL;DR

This work tackles the dual challenges of efficiency and temporal dynamics in sequential recommendation by proposing TALE, a temporal linear item-item model. TALE inserts temporal context through single-target augmentation, time interval-aware weighting, and trend-aware normalization, while retaining a closed-form training solution for fast learning and inference. Across five benchmark datasets, TALE matches or outperforms ten baselines—including neural SR models—achieving up to 18.71% accuracy gains and up to 30.45% improvements on long-tail items, with substantial speedups over neural approaches. The approach demonstrates practical scalability for temporally evolving user preferences, offering a strong, efficient alternative to complex neural architectures in SR tasks.

Abstract

In sequential recommendation (SR), neural models have been actively explored due to their remarkable performance, but they suffer from inefficiency inherent to their complexity. On the other hand, linear SR models exhibit high efficiency and achieve competitive or superior accuracy compared to neural models. However, they solely deal with the sequential order of items (i.e., sequential information) and overlook the actual timestamp (i.e., temporal information). It is limited to effectively capturing various user preference drifts over time. To address this issue, we propose a novel linear SR model, named TemporAl LinEar item-item model (TALE), incorporating temporal information while preserving training/inference efficiency, with three key components. (i) Single-target augmentation concentrates on a single target item, enabling us to learn the temporal correlation for the target item. (ii) Time interval-aware weighting utilizes the actual timestamp to discern the item correlation depending on time intervals. (iii) Trend-aware normalization reflects the dynamic shift of item popularity over time. Our empirical studies show that TALE outperforms ten competing SR models by up to 18.71% gains on five benchmark datasets. It also exhibits remarkable effectiveness in evaluating long-tail items by up to 30.45% gains. The source code is available at https://github.com/psm1206/TALE.

Figures (10)

• Figure 1: An example of the user sequence with different time intervals on the Beauty dataset.
• Figure 2: Co-occurrences between two consecutive items over average time intervals on the ML-1M and Beauty datasets. Appendix \ref{['appen:intv_co_occur']} shows a similar trend on Toys, Sports, and Yelp.
• Figure 3: Illustration of three components of TALE. Single-target augmentation takes each item in the input sequence as the target and the previous items as the source to form the training matrices. Time interval-aware weighting adjusts the significance of each source item, and Trend-aware normalization normalizes items based on their popularity around the time.
• Figure 4: Accuracy comparison by the time interval group on ML-1M and Beauty.
• Figure 5: Performance of TALE over $c$, $\tau_\text{time}$, and $N$ on ML-1M and Beauty.