This project investigates the factors influencing home ownership using Support Vector Machines (SVM) with linear, radial, and polynomial kernels. Based on data from the Integrated Public Use Microdata Series (IPUMS USA), this model aims to uncover how demographic and housing factors affect ownership status and provide insights for housing policy improvements.
- Predict home ownership status using SVM classifiers
- Compare kernel performances: Linear, Radial Basis Function (RBF), and Polynomial
- Identify top features influencing ownership (e.g., bedrooms, age, number of families)
- Explore social and economic implications of model findings
- Data Cleaning: Handled duplicates, irrelevant columns, and categorized key attributes
- Modeling:
- Linear Kernel: Grid search for optimal
C - Radial Kernel: Tuned
Candgamma - Polynomial Kernel: Tuned
Cand polynomialdegree
- Linear Kernel: Grid search for optimal
- Evaluation:
- Accuracy Scores: Linear (82.80%), Radial (83.48%), Polynomial (83.07%)
- Feature importance via coefficient magnitude and kernel sensitivity
- Decision boundary plots for each kernel type
| Kernel | Accuracy | Top Features |
|---|---|---|
| Linear | 82.80% | Bedrooms, Number of Families |
| Radial | 83.48% | Age, Bedrooms |
| Polynomial | 83.07% | Bedrooms, Age |
- The number of bedrooms consistently appeared as the most impactful feature across all kernels.
- Age and household size also play key roles in ownership likelihood.
- Larger households and younger individuals may face more barriers to home ownership.
- Findings suggest policy support could focus on first-time homebuyers and family-sized housing units.
- Python (Scikit-learn, Pandas, Matplotlib)
- SVM Kernels: Linear, RBF, Polynomial
- GridSearchCV for parameter tuning
- IPUMS USA Housing Dataset