STP is the generative model and inference algorithm for the users' behaviours in location-based social networks (LBSN). The MATLAB source codes and datasets can be used only for academic purposes. More at project website.
- The implementations of our method and the baselines.
- A dataset including the adjacency matrix and information of about 60000 checkins of 1000 users in Brazil.
- A collection of standard evaluation measures for the temporal and spatial predictions
- m-files for plotting the performance measures.
The project is executed successfully on Matlab R2015a. You also may need to install the Optimization toolbox beforehand.
To reproduce the results on synthetic data, run the exec_synth.m. For the real data, use the exec_real.m.
You can configure model parameters through those files as well. dataset_final.mat is the main data of the project and it consists of more than 60000 events of 1000 the Brazilian users and the corresponding adjacency matrix.
Results will be saved in Result folder.
For the edge recovery evaluation of algorithm using the AUC measure on synth data, you need to run exec_synth_recovery_test.m. The results will be saved in Results_unknown_adjacency folder.
Please note that this script works only if you perform the exec_synth.m with the option of unknown_adjacency beforehand.
The results are plotted with the code within the plot folder. Each m-file is associated with a single experiment results.