A Consumption-Generation Driven Place-Transition Embedding Framework for Predictive Process Monitoring

Installation

1. Create a python environment

   conda create -n PTE python=3.8.0
   conda activate PTE 

2. Install pytorch

   Following the official website's guidance (https://pytorch.org/get-started/locally/), install the corresponding PyTorch version based on your CUDA version. For our experiments, we use torch 1.12.1+cu116. The installation command is as follows:

   pip install torch==1.12.1+cu116 torchvision==0.13.1+cu116 torchaudio==0.12.1 --extra-index-url https://download.pytorch.org/whl/cu116

3. Install other related dependencies

   pip install -r requirements.txt

Train

First, you need to specify data_path and dataset in configs/PTE_Model.yaml.

Here, Two training methods are provided here:

1. Specify Hyperparameters: Specify model_parameters in configs/PTE_Model.yaml.

   python train_pte.py

2. Use Optuna for Hyperparameter Optimization:

   python run_pte.py

Test

```bash
python test_pte.py
```