Predict Company Bankruptcy using Machine Learning

Table of Contents

1. Project Motivation and Description
2. Proposed solution
3. Dataset
4. Technologies Used
5. Installation
6. Makefile Commands
7. Usage
8. Accessing the WebApp

Project Motivation and Description

In today's dynamic business environment, anticipating financial risks is crucial. This project aims to develop a predictive model for company bankruptcy using advanced machine learning algorithms and financial data analysis. By identifying early signs of financial distress, stakeholders can make informed decisions to prevent bankruptcy, minimizing financial losses.

We explore various machine learning models to predict company bankruptcy, focusing on improving accuracy. The dataset, from Polish companies, spans 2000-2013 and includes 64 features related to profitability, liquidity, solvency, and operational efficiency. Our project will explore and compare the usage of different machine learning models to predict company bankruptcy. We will evaluate the different models using various metrics, and improve their accuracy for better model performance.

Proposed solution

We employ multiple predictive models, including logistic regression, k-nearest neighbors, and decision trees, with logistic regression as the benchmark. Ensemble methods like bagging, boosting, and random forests will enhance predictive capabilities, providing a comprehensive understanding of financial risk.

Dataset

Source: Polish Companies Bankruptcy Data

Features Include:

• Net profit / total assets
• Total liabilities / total assets
• Working capital / total assets
• Current assets / short-term liabilities
• And many more financial ratios and metrics.

Installation

1. Clone the Repository

   git clone https://github.com/ChinSekYi/Bankruptcy-Prediction.git
   cd Bankruptcy-Prediction
   

2. Create and activate virtual environment
   Using Anaconda:

   • Miniconda Installation Guide
   • Windows Installation Guide

   conda create -n dev python=3.11 
   conda activate dev
   

   Using venv:

   python -m venv dev
   source dev/bin/activate  #On macOS/Linux
   source dev/bin/activate  #On Windows
   

3. Install Project Dependencies

   make install
   

Makefile Commands

• make install: Upgrades pip and installs the dependencies listed in requirements.txt.
• make test: Runs tests using pytest on test.py. (Note: You can also run tests on notebooks by uncommenting the #--nbval line if needed.)
• make format: Formats all Python files using isort and black.
• make run: Runs the main.py script.
• make lint: Lints the code with pylint, excluding certain message types (R and C) for the specified directories and files.
• make all: Executes install, format, and lint commands in sequence.

Usage

Open the following 3 notebooks:

• For Exploratory Data Analysis, Data Cleaning and Data Preprocessing:

  jupyter notebook notebook/EDA.ipynb   
  

• For Feature Selection:

  jupyter notebook notebook/FEATURE_SELECTION.ipynb   
  

• For Model Training and Evaluation:

  jupyter notebook notebook/MODEL_TRAINING.ipynb   
  

Accessing the WebApp

1. Run Flask app using CLI

   cd Bankruptcy-Prediction
   python3 app.py
   

   Then, access the application by clicking on one of the links below:

   • Home page
   • Prediction page

2. Access WebApp deployed through Microsoft Azure (Not active for now to avoid billing)

   • Home page
   • Prediction page