Qamar-usman-ai/SME-Financial-Health-Predictor

A high-performing classification pipeline predicting the Financial Health Index of Southern African SMEs. Achieved a 0.8849 F1-score (Rank 28) using a 800-tree Random Forest classifier driven by dense 30-dimensional Word2Vec categorical embeddings and robust missing data handling. # SME Financial Health Prediction System ## Predicting Financial Well-Being of Small Businesses in Southern Africa **Competition Score: 0.8849 28/900** --- ## 📋 Project Overview This project develops a machine learning system to predict the **Financial Health Index (FHI)** of Small and Medium-sized Enterprises (SMEs) across Southern Africa. The FHI classifies businesses into three categories: **Low**, **Medium**, or **High** financial health based on socio-economic, business, and financial data. ### Problem Statement Across Southern Africa, SMEs are crucial for employment and economic growth but face significant challenges: - Limited access to credit - Unstable cash flow - Exposure to economic shocks - Exclusion from formal financial systems Traditional metrics (revenue, profit) don't capture true financial health. This project provides a holistic measure reflecting: - **Savings and assets** - **Debt and repayment ability** - **Resilience to shocks** - **Access to credit and financial services** ### Dataset Information - **Source Countries**: Eswatini, Lesotho, Zimbabwe, Malawi - **Total Records**: 9,618 SME survey responses - **Features**: 39 (socio-economic, business, and financial indicators) - **Target Classes**: Low, Medium, High (multi-class classification) - **Missing Data**: 35 of 39 features contain missing values (range: 0.02% - 46.67%) --- ## 🎯 Model Performance ### Validation Metrics (10% Holdout) | Metric | Value | |--------|-------| | **Accuracy** | 88.88% | | **Weighted F1-Score** | 0.8858 | | **Macro F1-Score** | 0.8446 | ### Per-Class Performance | Class | Precision | Recall | F1-Score | Support | |-------|-----------|--------|----------|---------| | **High** | 0.89 | 0.72 | 0.80 | 47 | | **Low** | 0.90 | 0.96 | 0.93 | 628 | | **Medium** | 0.86 | 0.75 | 0.80 | 287 | ### Leaderboard Score - **Public Score**: 0.8849 - **Ranking**: Top 28 (among 950+ participants) --- ## 📁 Project Structure ``` SME-Financial-Health-Prediction/ ├── README …