Abstract:Next location recommendation plays an important role in various location-based services, yielding great value for both users and service providers. Existing methods usually model temporal dependencies with explicit time intervals or learn representation from customized point of interest (POI) graphs with rich context information to capture the sequential patterns among POIs. However, this problem is perceptibly complex because various factors, e . g ., users’ preferences, spatial locations, time contexts, activity category semantics, and temporal relations, need to be considered together, while most studies lack sufficient consideration of the collaborative signals. Toward this goal, we propose a novel M ulti-Level C ollaborative Neural N etwork for next location Rec ommendation (MCN4Rec). Specifically, we design a multi-level view representation learning with level-wise contrastive learning to collaboratively learn representation from local and global perspectives to capture complex heterogeneous relationships among user, POI, time, and activity categories. Then a causal encoder-decoder is applied to the learned representations of check-in sequences to recommend the next location. Extensive experiments on four real-world check-in mobility datasets demonstrate that our model significantly outperforms the existing state-of-the-art baselines for the next location recommendation. Ablation study further validates the benefits of the collaboration of the designed sub-modules. The source code is available at https://github.com/quai-mengxiang/MCN4Rec.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to more accurately recommend the next possible location for a user to visit in location - based services. Existing methods usually model temporal dependencies through explicit time intervals or learn representations from customized Point - of - Interest (POI) graphs with rich context information to capture the sequence patterns among POIs. However, this problem is very complex because multiple factors need to be comprehensively considered, such as user preferences, spatial locations, temporal contexts, activity - category semantics, and temporal relationships, and most studies lack sufficient consideration of these collaborative signals. In addition, existing methods also have data sparsity and cold - start problems, which seriously affect the accuracy of recommendations. To this end, the authors propose a multi - level collaborative neural network model named MCN4Rec, aiming to solve the problems of data sparsity, cold - start, and complex association learning. The model improves the recommendation performance in the following ways: 1. **Multi - level view representation learning**: Construct a knowledge graph to model various relationships among users, POIs, and activity categories, and learn their representation vectors from global and local perspectives through different embedding techniques. 2. **Hierarchical contrastive representation learning**: Use local - level and global - level contrastive learning to constrain node representation learning between different local views and between global and local views to maximize the consistency of representations among different views. 3. **Category - aware temporal representation**: Use time2vec to learn time periodicity and incorporate activity - category semantics into the temporal representation. 4. **Sequence modeling module**: Utilize a causal encoder to further learn the sequence movement patterns of the learned check - in sequence representations to recommend the next location. Through the above methods, the MCN4Rec model has been extensively experimented on four real - world mobile datasets. The results show that the model significantly outperforms the existing state - of - the - art baseline models, with an average improvement of 11.70% and 8.28% in the Acc@1 and MRR metrics respectively. In addition, ablation studies further verify the advantages of the design sub - module collaboration. The model also demonstrates the ability to handle the cold - start problems of short - trajectory and inactive - user check - in data.

MCN4Rec: Multi-Level Collaborative Neural Network for Next Location Recommendation

Next Point-Of-Interest Recommendation With Temporal And Multi-Level Context Attention

Attentive Sequential Model Based on Graph Neural Network for Next Poi Recommendation.

A Multi-Channel Next POI Recommendation Framework with Multi-Granularity Check-in Signals

An improved deep sequential model for context-aware POI recommendation

NEXT: a neural network framework for next POI recommendation

Next POI Recommendation Based on Location Interest Mining with Recurrent Neural Networks

Empowering Next POI Recommendation with Multi-Relational Modeling

Intent-aware Graph Neural Network for Point-of-Interest Embedding and Recommendation

STA-TCN: Spatial-temporal Attention over Temporal Convolutional Network for Next Point-of-interest Recommendation

MST-RNN: A Multi-Dimension Spatiotemporal Recurrent Neural Networks for Recommending the Next Point of Interest

MNN4Rec: A relation-aware approach based on multi-view news network for news recommendation

DAN-SNR: A Deep Attentive Network for Social-Aware Next Point-of-Interest Recommendation

Attention mechanism and adaptive convolution actuated fusion network for next POI recommendation

Where to Go Next: A Spatio-Temporal Gated Network for Next POI Recommendation

A Spatio-temporal Adaptive Personalized Meta-recommender for Next Location

Next Point-of-Interest Recommendation with Auto-Correlation Enhanced Multi-Modal Transformer Network

Context-aware multi-head self-attentional neural network model for next location prediction

Kernel-based Substructure Exploration for Next POI Recommendation

Cross-Task Multimodal Reinforcement for Long Tail Next POI Recommendation

Multi-granularity contrastive learning model for next POI recommendation