knockouts-redesign

About

This is the implementation of a data-driven approach for discovering improvement opportunities related to the execution of knock-out checks within business processes.

The tool comes with a browser-based UI (powered by Streamlit), but it can also be invoked via CLI or imported as a module in Python scripts.

Resources

• Download the full paper here
• Download the full thesis here
• Event logs used for the paper experiments
• Results of baseline experiments

Index

1. Set-up
2. Usage
3. Running tests
4. Notes

Set-up

Option A: Using a setup script

• Setup scripts are provided for Windows & Linux, which perform the steps described in the "Manual steps" section.
• Requirements:
  • Having cloned the repository
  • Python 3.9+ installed
• Commands:

  # Windows: 
  setup.bat
       
  # Linux: 
  bash setup.sh

Option B: Manual steps

Expand

• Clone the repo

• Create a virtual environment with venv

• In the directory that contains the newly created venv folder, run:

  source venv/Scripts/activate
  pip install -r requirements.txt
  

• Depending on your platform, install graphviz (v 3.0.0+) separately, and make sure it's in your PATH (see also: graphviz downloads page):

   # Windows: 
   choco install graphviz
   
   # Linux: 
   sudo apt install graphviz
   
   # Mac:
   brew install graphviz

Usage

Create a config file following the schema defined in config_schema.json (the directory also contains examples). Then you can choose whether to:

• Launch the tool as a web app (powered by Streamlit):

     # Windows: 
     start.bat
     
     # Linux: 
     bash start.sh

• Launch from CLI:

     # Windows:
     python main.py config/synthetic_example.json
  
     # Linux: 
     python3 main.py config/synthetic_example.json

• Or programmatically:

  from knockout_ios.pipeline_wrapper import Pipeline
  
  ko_redesign_adviser = Pipeline(config_file_name="synthetic_example.json").run_pipeline()

  In all these cases, if using the default settings, output will be written to .csv and txt files in the data/outputs folder.

Running tests

Shell scripts are provided, which run the tests in parallel (thanks to pytest-xdist).

 # Windows: 
 test.bat
 
 # Linux: 
 bash test.sh

Notes

About wittgenstein & grid search

Expand

• Excessive warnings are printed due to frame.append deprecation warnings in the wittgenstein package.

• Currently, they can just be suppressed by passing the -Wignore flag.

• Another situation where excessive warnings are printed, is during the parameter grid search, as some combinations may yield too small training sets or sets with no positive examples.

• This can be safely ignored as long as the grid contains at least one parameter combination that results in a valid dataset, and the rule discovery module is able to return meaningful rules.