Abstract:Aiming at the problem that the existing sequence recommendation model ignores the personalized behavior of different users,the model cannot fully capture the interest drift caused by users’ dynamic preferences,a sequence recommendation model based on users’ long and short term preferences(ULSP-SRM)is proposed.Firstly,the dynamic category embedding of the user is generated according to the category and time information of the interactive items in the user’s sequence,thereby effectively establishing the correlation between the items and reducing the sparsity of the data.Secondly,according to the time interval information of the user’s current clicked item and the last clicked item,a personalized time series position embedding matrix is generated to simulate the user’s personalized aggregation phenomenon and better reflect the dynamic change of user preference.Then,the user’s long-term preference sequence fused with the personalized time-series position embedding matrix is input into the gated recurrent unit in units of sessions to generate the user’s long-term preference representation,and the user’s long and short term preferences are fused through the attention mechanism to generate the final preference representation of the user,to achieve the purpose of fully capturing the user’s preference.Finally,the final user preference representation is input to the recommendation prediction layer for the next recommendation prediction.Experiments are carried out on seven subsets of Amazon public data set,and the area under curve(AUC ),recall rate and precision rate indicators are used for comprehensive evaluation.Experimental results show that the proposed model outperforms other advanced benchmark models,effectively improving recommended perfor-mance.

Time-Aware Preference Recommendation Based on Behavior Sequence

Long- and Short- Term Sequential Recommendation with Temporal Interval.

Time Matters: Sequential Recommendation with Complex Temporal Information

Time-aware Sequence Model for Next-Item Recommendation.

An Improved Sequential Recommendation Algorithm based on Short-Sequence Enhancement and Temporal Self-Attention Mechanism

Sequential Recommendation Model Based on User's Long and Short Term Preference

Time Interval Aware Self-Attention for Sequential Recommendation

Fusing User Preferences and Spatiotemporal Information for Sequential Recommendation

Enhancing Repeat-Aware Recommendation from a Temporal-Sequential Perspective

Personalized Behavior-Aware Transformer for Multi-Behavior Sequential Recommendation

Attention with Long-Term Interval-Based Deep Sequential Learning for Recommendation.

Learn over Past, Evolve for Future: Search-based Time-aware Recommendation with Sequential Behavior Data

Multi-behavior collaborative contrastive learning for sequential recommendation

Long Short-Term Preference Modeling for Continuous-Time Sequential Recommendation

Recognizing Semantics-Consistent Subsequences for Sequential Recommendation

Continuous-Time Sequential Recommendation with Temporal Graph Collaborative Transformer

Long and Short-Term User Intent Learning for Sequence Recommendation

Multi-Behavior Sequential Recommendation with Temporal Graph Transformer

Sequential Recommender via Time-aware Attentive Memory Network

Attention with Long-Term Interval-Based Gated Recurrent Units for Modeling Sequential User Behaviors

Déjà vu: A Contextualized Temporal Attention Mechanism for Sequential Recommendation