Bankruptcy Prediction in Poland

Overview

This project aims to predict the likelihood of bankruptcy for companies in Poland using historical financial data. The project utilizes machine learning models to assess a company’s financial health based on various financial ratios and features. The goal is to provide insights that can help companies and stakeholders identify financial risks and take preventive measures.

Project Highlights

• Objective: Build a predictive model that assesses a company’s bankruptcy risk using historical financial data.
• Data Source: Dataset collected by Polish economists, containing financial information of companies over several years.
• Tools Used: Python, Pandas, scikit-learn, matplotlib, seaborn, and Linux command line.
• Techniques: Logistic regression, resampling techniques, decision trees, classification metrics, and data wrangling.

Workflow

1. Data Collection:

   • Accessed a publicly available dataset from Polish economists studying corporate bankruptcy.
   • Loaded and processed the data using the Linux command line and Python.

2. Data Preprocessing:

   • Performed extensive data cleaning using Pandas, addressing missing values and outliers.
   • Handled imbalanced data using oversampling techniques to ensure that the model could accurately predict minority cases of bankruptcy.
   • Conducted exploratory data analysis (EDA) to identify key financial ratios and features that influence bankruptcy risk.

3. Modeling:

   • Built multiple classification models, including logistic regression and decision tree classifiers, to predict bankruptcy risk.
   • Fine-tuned model hyperparameters using grid search and cross-validation to optimize performance.
   • Evaluated models using classification metrics such as accuracy, precision, recall, and F1-score to determine the best-performing model.

4. Prediction:

   • Deployed the best-performing model to predict the likelihood of bankruptcy for new companies.
   • Visualized prediction results to highlight the most critical factors influencing bankruptcy.

Results

• The logistic regression model performed well in predicting bankruptcy, with high precision and recall for the positive (bankruptcy) class.
• The model was able to capture significant financial trends and patterns that indicate a company’s potential risk of bankruptcy.
• The use of resampling techniques helped in effectively handling the imbalanced nature of the dataset.

Key Learnings

• Financial Analysis: Gained deeper insights into how financial ratios and key indicators can affect a company's bankruptcy risk.
• Data Imbalance: Successfully handled imbalanced datasets using resampling techniques to ensure that the minority class (bankruptcy cases) was adequately represented in the model.
• Machine Learning: Improved skills in building and optimizing classification models for real-world financial problems, with a focus on logistic regression and decision trees.

Future Improvements

• Incorporate Time-Series Data: Add historical data trends and time-series analysis to improve the model’s ability to predict long-term bankruptcy risks.
• Advanced Ensemble Models: Experiment with advanced ensemble models such as Random Forest, Gradient Boosting, or XGBoost to improve prediction accuracy.
• Interactive Dashboard: Develop a user-friendly dashboard that allows businesses to input their financial data and receive real-time bankruptcy risk predictions.

Installation and Setup

1. Clone this repository:

   git clone https://github.com/nisha2k21/bankruptcy-prediction-poland.git

2. Navigate to the project directory:

   cd bankruptcy-prediction-poland

3. Install the required dependencies:

   pip install -r requirements.txt

4. Run the model to make predictions:

   python src/bankruptcy_prediction.py

Project Structure

├── data
│   └── bankruptcy_data.csv          # Raw data file
├── notebooks
│   └── EDA_and_Modeling.ipynb       # Jupyter notebook for analysis and model building
├── src
│   └── bankruptcy_prediction.py     # Python script for running the model
├── requirements.txt                 # List of required dependencies
└── README.md                        # Project documentation

Technologies Used

• Programming Language: Python
• Libraries: Pandas, NumPy, scikit-learn, matplotlib, seaborn
• Machine Learning: Logistic regression, Decision tree classifiers, Resampling techniques
• Model Evaluation: Precision, Recall, F1-score, Accuracy
• Platform: Linux command line for data handling

Contact

For any questions, feedback, or collaboration opportunities, feel free to reach out:

• Email: [email protected]
• LinkedIn: Nisha Kumari