Corporate Valuation & Bankruptcy Risk Prediction Dashboard

Korean Listed Company Financial Analysis Platform
An end-to-end financial analysis system covering 42 KOSPI-listed companies,
built on DART public filings and Yahoo Finance data.

---

Project Overview

This project replicates a buy-side investment analysis workflow from scratch.
It collects and processes public financial data, applies the Altman Z-Score model
alongside machine learning classifiers, and delivers an interactive dashboard
for assessing corporate financial health.

Problem Statement

"Can we systematically identify bankruptcy risk from financial statement data alone?"

Evaluating the financial soundness of investee companies is a core part of the investment process.
However, manually analyzing dozens of firms across multiple years is time-consuming and error-prone.
This project automates and visualizes that process end-to-end.

---

Tech Stack

Category	Tools
Data Collection	DART OpenAPI, yfinance
Data Processing	Python, pandas, numpy
Machine Learning	scikit-learn (Random Forest, Gradient Boosting)
Visualization	Streamlit, Plotly
Environment	python-dotenv

---

Project Structure

valuation/
├── config.py              # Company universe (42 firms) & API config
├── dart_collector.py      # DART financial statement collection & processing
├── price_collector.py     # Stock price collection & annual metrics
├── valuation_model.py     # Altman Z-Score + ML bankruptcy prediction
├── dashboard.py           # Streamlit dashboard (5 pages)
├── .env                   # API credentials (excluded from Git)
└── data/
    ├── raw/               # Raw collected data
    ├── processed/         # Cleaned master dataset
    └── model/             # Model outputs & scorecards

---

Methodology

1. Altman Z-Score (1968)

A classic bankruptcy prediction model combining five financial ratios.

Z = 1.2×X1 + 1.4×X2 + 3.3×X3 + 0.6×X4 + 1.0×X5

Variable	Description	Proxy
X1	Liquidity	Working Capital / Total Assets
X2	Cumulative Profitability	Net Income / Total Assets
X3	Operating Efficiency	EBIT / Total Assets
X4	Financial Leverage	Equity / Total Liabilities
X5	Asset Utilization	Revenue / Total Assets

Classification Thresholds

Z-Score	Zone
Z > 2.99	🟢 Safe Zone
1.81 < Z < 2.99	🟡 Grey Zone
Z < 1.81	🔴 Distress Zone

2. ML Bankruptcy Prediction

Implemented to complement the Z-Score model's limitations.

• Algorithms: Random Forest, Gradient Boosting (best model selected by F1)
• Features: 13 financial indicators spanning profitability, stability, growth, and stock volatility
• Validation: Stratified 5-Fold Cross Validation
• Labeling: A firm is labeled as high-risk if it meets 3 or more of the following conditions:
  • Z-Score < 1.81
  • Debt-to-Equity > 200%
  • Current Ratio < 100%
  • Operating Margin < 0%
  • Operating Cash Flow < 0

3. Model Performance

Model	F1 Score	AUC
Random Forest	0.400 ± 0.490	0.971 ± 0.042
Gradient Boosting	0.467 ± 0.400	0.886 ± 0.194

The Random Forest classifier achieved an AUC of 0.971 in Stratified 5-Fold cross-validation, demonstrating strong discriminative power in identifying financially distressed firms.
The lower F1 score is expected given class imbalance (few distressed firms in the dataset), while AUC more accurately reflects the model's ranking ability across all thresholds.

---

Data Coverage

• Universe: 42 KOSPI-listed companies (sector representatives)
• Period: FY2019 – FY2023 (5 years)
• Financial Statements: Consolidated (IFRS)
• Sectors: Semiconductors/IT · Automotive/Battery · Chemicals/Energy · Bio/Healthcare · Consumer/Retail · Telecom/Media · Construction

---

Dashboard Features

Page	Content
Overview	KPI summary, Z-Score distribution, profitability vs. stability bubble chart
Z-Score Ranking	Full ranking table, Z-Score component decomposition chart
Financial Trends	Multi-company time series, 4-metric simultaneous comparison
Sector Comparison	Sector box plots, radar chart
Company Scorecard	Individual company deep-dive, 5-year financial summary

---

Getting Started

1. Install Dependencies

pip install pandas numpy scikit-learn streamlit plotly yfinance python-dotenv lxml requests

2. Set Up API Key

Obtain a free API key from DART OpenAPI and create a .env file:

DART_API_KEY=your_api_key_here

3. Collect Data

python dart_collector.py    # Financial statements (~10 min)
python price_collector.py   # Stock prices & merge

4. Run the Model

python valuation_model.py

5. Launch the Dashboard

streamlit run dashboard.py

---

Limitations & Future Work

Known Limitations

1. Industry Bias in Altman Z-Score
The original model was designed for manufacturing SMEs in the 1960s.
For large capital-intensive firms (e.g., semiconductors) or asset-light tech platforms,
X4 (Equity/Total Liabilities) tends to be disproportionately large, inflating the Z-Score.
This explains why fundamentally sound companies like SK Hynix and NAVER
are classified in the distress zone despite strong underlying business performance.

2. Synthetic Labeling
In the absence of actual bankruptcy event data for Korean listed companies,
distress labels were generated using rule-based financial thresholds.
Class imbalance (few distressed firms) causes the ML model to predict conservatively, resulting in lower F1 scores despite high AUC.

3. Financial Sector Exclusion
Banks, securities firms, and insurers were excluded due to structural differences
in their financial statements. Analyzing these firms requires sector-specific metrics
such as BIS capital ratios and net interest margins (NIM).

Future Improvements

• Apply Zmijewski Model or Ohlson O-Score for multi-model comparison
• Implement sector-specific thresholds to account for industry characteristics
• Incorporate actual delisting / workout event data for supervised labeling
• Extend to quarterly data for higher-frequency monitoring
• Add a DCF Valuation module for intrinsic value estimation

---

Key Insights

Samsung Electronics — Z-Score: 3.345
The majority of the Z-Score (X4 contribution: 2.36) stems from its exceptionally low leverage,
with a debt-to-equity ratio of approximately 26%.
This suggests Samsung's financial safety is driven by its balance sheet strength
rather than operating profitability alone.

GS Engineering & Construction — Bankruptcy Probability: 100%
Classified as high-risk by both the Z-Score (0.928) and the ML model.
A combination of rising leverage and deteriorating operating cash flow
points to structural financial stress, consistent with publicly reported challenges.

---

References

• Altman, E. I. (1968). Financial Ratios, Discriminant Analysis and the Prediction of Corporate Bankruptcy. Journal of Finance, 23(4), 589–609.
• DART OpenAPI Documentation: https://opendart.fss.or.kr

---

License

MIT License © 2026 hazel-jeon