Decoding Salaries – Predictive Analysis of Job Market Trends
Domain & Tools
Job Market Intelligence / Salary Prediction Python, Pandas, Scikit‑learn, Seaborn, Google Colab
Techniques
- Linear Regression, Ridge, Lasso, ElasticNet
- VIF Analysis for multicollinearity
- Residual diagnostics & model stability checks
Overview
Conducted a deep‑dive regression analysis to predict salaries from job descriptions and company traits. Explored multiple linear models and revealed very weak linear relationships across features, emphasizing the need for richer feature engineering.
Highlights
- Cleaned and transformed salary data (hourly → annual).
- Created composite features like Experience × Rating and Company Size × Revenue.
- Identified multicollinearity using VIF; flagged Revenue features with VIF > 25.
- Applied ElasticNet for balanced prediction and interpretability.
- Delivered strategic insights for HR teams on compensation benchmarking.
Key Insight Summary
Salary prediction showed weak linear relationships, revealing that job market compensation requires deeper feature engineering, richer text extraction, and non‑linear modeling for real‑world accuracy.
← Back to Home