StockWell: Measuring the Financial Wellness of S&P 500 Companies

KWK Machine Learning x Goldman Sachs — Final Project By: Hailey Muñiz

📌 Overview

StockWell is a machine learning project designed to evaluate the financial stability of S&P 500 companies using publicly available financial metrics.

The goal is to create a simple, transparent Financial Health Score (FHS) that anyone can use — students, new investors, or anyone curious about company stability — without relying on private or proprietary scoring systems.

Research Question

Can we use machine learning to predict a company’s near-future financial health using a few key indicators such as:

• EBITDA (Earnings Before Interest, Taxes, Depreciation, and Amortization)
• Revenue Growth
• Current Price
• Market Capitalization

Dataset

Source: S&P 500 Stocks (Larxel — Kaggle Dataset) This dataset includes financial and company information for all S&P 500 companies.

Methods & Workflow

1. Data Cleaning

• Filled missing values using the median
• Removed non-essential columns
• Converted categorical variables (sector, industry) into numeric form
• Standardized numeric features using Z-score scaling

2. Feature Engineering — Financial Health Score (FHS)

Created a custom score inspired by the Altman Z-Score:

FHS = Z(Ebitda) + Z(RevenueGrowth) + Z(CurrentPrice) + Z(MarketCap)

Higher FHS → better financial health.

3. Modeling — Random Forest Regression

Trained 3 models with different depths + tree counts to predict next-period FHS.

4. Evaluation Metrics

• MAE: ~0.88–0.92
• RMSE: ~2.15
• R²: ~0.55

Model 1 (100 trees, depth 4) performed best.

⭐ Key Findings

1️Market Capitalization dominates

It was the strongest predictor (90%+ feature importance). → Larger companies tend to be more financially stable.

2️Short-term indicators matter less

Revenue growth, EBITDA, and stock price had much smaller influence.

3️Simple model performed best

More complex models (deeper trees, more estimators) did not improve performance.

Limitations

• Only uses four financial metrics
• Only includes large S&P 500 companies
• Small dataset → limits model complexity
• Market cap may overly dominate predictions
• No historical time series data

Future Improvements

• Add financial ratios (debt, liquidity, profitability)
• Use historical data instead of single snapshots
• Try advanced models: Gradient Boosting, XGBoost, Neural Networks
• Build scenario analysis tools (e.g., recession, interest rate changes)

Files in This Repository

• StockWell_Measuring_the_Financial_Wellness_of_S&P_500_Companies_(_KWK_Machine_Learning_x_Finance_Challenge)_Final_Project.ipynb – Main notebook with code + results
• StockWell Measuring the Financial Wellness of the S&P 500 (KWK Machine Learning x Finance Challenge).pdf – Slide deck for the project

Video Presentation

▶️ Watch the Presentation: https://www.tella.tv/video/stockwell-sandp-500-financial-health-scoring-3-bnoo

Acknowledgements

This project was completed for the KWK Machine Learning x Goldman Sachs. Special thanks to mentors, instructors, and open-source resources used throughout the process.