• Getting Started
• Requirements
• Running the code
• Authors
  • Adam Broniewski
  • Himanshu Choudhary
  • Tejaswini Dhupad
• License

This project contains the technical part of Predicting Rejected Declarations in the Travel Management System: BPI data challenge 2020

• Python 3.9
• pandas
• sklearn
• numpy
• matplotlib
• plotly
• pm4py
• seaborn
• tabulate

These steps will run the full data-preparation, model building, prediction generations.

1. Install the dependencies:

   pip install -r requirments.txt

   Note: The scripts have been developed and tested with Python 3.9

   Note 2: If using ARM32 or ARM64 machine and an error is encountered installing pm4py, follow pm4py install instructions here.

2. Once all dependencies are installed, we can run the data preperation python file (default length 6 with complex encoding). The outputs are saved in training_data/encodings directory.

   python3 src/data_preparation.py

   Expected Outcome -

   train test and validation split times are - 2018-09-17 11:34:42 , 2018-11-08 15:42:36 
   train, val and test count
   4020 861 861
   preparing training data for prefix length 6
   preparing test data for prefix length 6
   preparing val data for prefix length 6
   Done!
   

3. The below code will run our machine learning model (Default Decision Trees) on generated encodings. The outputs are saved in results/ directory.

   python3 src/model_building_and_evaluation.py

   Expected Outcome -

APM_BPI_2020 is open source software [licensed as MIT][license].