Time-Aware-Location-Prediction

This work "Time-Aware Location Prediction by Convolutional Area-of-Interest Modeling and Memory-Augmented Attentive LSTM" has been published in TKDE 2022.

📄 Description

t-LocPred is a novel time-aware location prediction model for Point of Interests (POIs) recommendation. It consists of a convolutional AoI modeling module (ConvAOI) and memory-augmented attentive LSTM (mem-attLSTM). It captures both coarse- and fine-grained spatiotemporal correlations among a user’s historical check-ins and models his/her long-term movement patterns.

🔧 Dependencies

• Python == 3.7 (Recommend to use Anaconda or Miniconda)
• PyTorch == 1.8.1
• NVIDIA GPU (RTX 3090) + CUDA 11.1

Installation

1. Clone repo

   git clone https://github.com/BIT-MCS/Time-Aware-Location-Prediction.git
   cd Time-Aware-Location-Prediction

2. Install dependent packages

   pip install -r requirements.txt
   

💿 Data Preparation

1. You can modify the config file one_day_data_general/conf.py for data preparation. For example, you can control the length of check-in sequences by modifying this line:

   [41]  'seq_len': 8,
   

2. Using the following commands to process the original datasets and generalize the data for t-LocPred.

   cd one_day_data_general
   python main.py

💻 Training

We provide complete training codes for t-LocPred.
You could adapt it to your own needs.

1. If you don't have NVIDIA RTX 3090, you should comment these two lines in file utils.py.

   [19]  torch.backends.cuda.matmul.allow_tf32 = False
   [20]  torch.backends.cudnn.allow_tf32 = False
   

2. You can modify the config files step1model/ConvAOI/conf.py.
   For example, you can control the hyperparameter about CNN kernal size in convolutional AoI modeling module by modifying this line:

   [31]  'cnn_kernel_size': 3,
   

3. Training the ConvAOI module:

   cd step1model/ConvAOI
   python main.py
   

4. Training the mem-attLSTM module:

   cd step2model/mem-attLSTM
   python main.py
   

🏁 Testing

1. Running the ConvAOI module:

   cd step1model/ConvAOI
   python test.py
   

2. Running the mem-attLSTM module:

   cd step2model/mem-attLSTM
   python test.py
   

📧 Contact

If you have any question, please email [email protected].

Paper

If you are interested in our work, please cite our paper as

@ARTICLE{9128016,
  author={Liu, Chi Harold and Wang, Yu and Piao, Chengzhe and Dai, Zipeng and Yuan, Ye and Wang, Guoren and Wu, Dapeng},
  journal={IEEE Transactions on Knowledge and Data Engineering}, 
  title={Time-Aware Location Prediction by Convolutional Area-of-Interest Modeling and Memory-Augmented Attentive LSTM}, 
  year={2022},
  volume={34},
  number={5},
  pages={2472-2484},
  doi={10.1109/TKDE.2020.3005735}
}