Introduction
-
The proposed model uses an approach based on a probabilistic weighting strategy using eleven graphs to tackle the sparsity problem.
-
Presents two algorithms to get users' favorite season(s) with the most visited categories in a particular season using past check-ins history.
-
The proposed approach uses the work of RELINE (Recommendation with Multiple Network Embeddings) [10] and tries to learn the embeddings of graphs by using the concept of graphs to find the heterogeneous preferences of users.
Literature review
Participated networks in the recommendation model
Symbol | Description |
---|---|
\({c}_{u,i}\) | User’s check-ins |
\(\Delta S\) | Season |
L | Location |
K | Location Category |
W | List of weights \({w}_{i,j}\) on each graph |
Point-of-Interest
Check-in
Season
User-user graph
User-category graph
User-season graph
User-location graph
Category
Category-location graph
Category-user graph
Category-category graph
Category-season graph
Location-location graph
Location-user graph
Location-season graph
Problem definition
Proposed next-POI recommender system
Learning embeddings for large information networks
Optimization of the model
Sample edge | \(O(1)\) |
Optimization of negative sampling | \(O\left(N+1\right)\) |
Total complexity | \(O(N\times E)\) |
Learning graph dynamics
Personalized next-POI recommendation
Employment of cloud and edge computing
Results and discussion
Dataset
POIs | 1048000 |
Check-ins | 2145800 |
Friendships | 607300 |
Baseline models
-
RELINE [10]: They have used a graph-based approach to learn users’ and POI relationships from 8 weighted networks in a hidden space and provide location recommendations under a strategy having a probability that examines the influence of social, geographical, temporal, and preference dynamics.
-
GE [28]: is another graph-based embedding model that exploits geographical influence, sequential effect, temporal cyclic effect, and semantic effect in a unified way and embeds four information graphs into a shared embedding space. Also, a novel time-decay method is proposed that dynamically computes the user’s latest preferences based on the embedding of his/her checked-in POIs learned in the embedding space.
-
WWO [31]: is a unified POI recommender system with temporal interval assessment that considers temporal interval distributions and developed the low-rank network model, identifying a set of bi-weighted network bases to learn the static preferences and dynamic preferences coherently.
-
PGB [33]: This probabilistic model employs the graph-based Markov chain method to improve recommendation accuracy. The choice of suggesting an item is conditioned by considering recommendations generated in previous steps.
Experiments evaluation
Impact of information graphs
Model | 1 | 5 | 10 | 15 | 20 |
---|---|---|---|---|---|
PGB | 0.21 | 0.24 | 0.32 | 0.33 | 0.43 |
GE | 0.25 | 0.32 | 0.37 | 0.40 | 0.42 |
WWO | 0.22 | 0.28 | 0.34 | 0.34 | 0.44 |
RELINE | 0.28 | 0.32 | 0.37 | 0.44 | 0.46 |
NPR-LBN | 0.29 | 0.31 | 0.38 | 0.45 | 0.49 |
Comparative analysis
Cold start user
Cold start locations
Significance of seasons
\(\Delta {\varvec{S}}\) | Acc@n | ||||
---|---|---|---|---|---|
5 | 0.27 | 0.35 | 0.24 | 0.25 | 0.32 |
10 | 0.23 | 0.32 | 0.42 | 0.31 | 0.32 |
15 | 0.32 | 0.38 | 0.39 | 0.42 | 0.45 |
20 | 0.33 | 0.42 | 0.23 | 0.98 | 0.32 |