Skip to content

Latest commit

 

History

History
140 lines (95 loc) · 4.84 KB

README.md

File metadata and controls

140 lines (95 loc) · 4.84 KB

Automated Generation of Process Simulation Scenarios from Declarative Control-Flow Changes

The code here presented is able to execute different pre- and post-processing methods and architectures for building and using generative models from event logs in XES format using LSTM anf GRU neural networks. This code can perform the next tasks:

  • Training LSTM neuronal networks using an event log as input.
  • Generate full event logs using a trained LSTM neuronal network.
  • Generate traces that adhere to a corresponding set of rules given by the user.
  • Predict the remaining time and the continuation (suffix) of an incomplete business process trace.
  • Discover the stochastic process model and generate a simulation based on the rules given by the user.

Architecture

alt text

System Requirements

  • Python 3.x
  • Java SDK 1.8 Choose right version according with Operative System.
  • Anaconda Distribution
  • Git

Getting Started

These instructions will get you a copy of the project up and running on your local machine for development and testing purposes.

git clone --recurse-submodules https://github.com/AdaptiveBProcess/DeclarativeProcessSimulation.git

Once the repository has been cloned, you can update the submodules at any time using the following command:

git submodule update --init --recursive

For Simod and GenerativeLSTM it is necessary to update version with this commands

cd GenerativeLSTM 
git checkout Declarative-Process
cd ..
Generative LSTM branch


cd Simod-2.3.1
git checkout v2.3.1
cd ..

Prerequisites

To execute this code with the previous Anaconda install in your system, create an environment using the environment.yml specification provided in the repository.

cd GenerativeLSTM
conda env create -f environment.yml
conda activate deep_generator

In GenerativeLSTM\output_files please create two folders GenerativeLSTM\output_files\simulation_files and GenerativeLSTM\output_files\simulation_stats, this is for the correct creation of the model

Be sure when running this script to be using Conda prompt or to configure conda into another prompt

Here is an example here for adding conda to Windows prompt in vs-code if needed https://stackoverflow.com/questions/54828713/working-with-anaconda-in-visual-studio-code

Running the script

To choose event-log the log must be in GenerativeLSTM\input_files line 43 on dg_training.py must be modified with the name of the event log, as the other parameters as model_family or opt_methods. After this training is done, a folder with this format "YYYYMMDD_XXXXXXXX_XXXX_XXXX_XXXX_XXXXXXXXXXX" is generated on this folder GenerativeLSTM\output_files On dg_prediction.py line 140 with

 parameters['filename'] = 'example.csv'

Change the name to the event log previously used on training on dg_prediction.py

Training the model

Train the model with the input event log.

python dg_training.py

This generates a folder in output_files. Copy that folder name into dg_prediction.py and replace the value of the variable parameters['folder'].

Generate predictions

For this step Java 1.8 SDK is needed.

Train the model with the input event log.

Before starting this step in the route \GenerativeLSTM\output_files please create a folder named \GenerativeLSTM\output_files\simulation_files for the program to find the exact route.

Archive GenerativeLSTM\rules.ini must be modified to make the rules correctly

Rules have this structure and must be written with this format

Rules specification

#- Directly folllows: A >> B

#- Eventually folllows: A >> * >> B

#- Task required: A

#- Task not allowed: ^A

Here is an example for rules.ini

# Production
[RULES]
path =  Validar solicitud >> Radicar Solicitud Homologacion
variation = =1

On line 159 on dg_prediction.py this parameter must be modified with the folder that was generated after the training.

parameters['folder'] = 'YYYYMMDD_XXXXXXXX_XXXX_XXXX_XXXX_XXXXXXXXXXX'

When training the model, be sure to use the appropiate rules related to the BPMN model that is being used. Rules.ini gives an idea of which rules can be used to ensure simulation data goes as well as posible

python dg_prediction.py

This generates a simulation process model corresponding to the implementation of the changes proposed by the user. The simulation files are stored in output_files/simulation_files/. In addition, the approach simulates that simulation process model and generates a statistics file corresponding to the stats of the simulated model. The simulation stats are stored in output_files/simulation_stats/.

Examples

The files used for the experimentation are stored in input_files.

Authors

  • Daniel Baron
  • Manuel Camargo
  • Marlon Dumas
  • Oscar Gonzalez-Rojas