HR Attrition Prediction Model

An enterprise with 5,000+ employees was experiencing an unusually high turnover rate in its R&D department. The 'cost of replacement' was estimated at 1.5x the annual salary per employee, causing significant financial and intellectual property loss.

HR exit interviews provided qualitative data, but the company needed a quantitative way to predict which employees were at risk *before* they resigned. They lacked a unified view of engagement and performance metrics across the organization.

I developed a Machine Learning model using the XGBoost algorithm to predict employee attrition risk. The feature set included over 50 variables, including engagement scores, training history, salary benchmarks, and manager feedback sentiment.

I integrated the model predictions into a secure HR dashboard, allowing managers to see 'Risk Scores' for their teams. I also implemented a 'What-If' analysis tool where HR could simulate the impact of different retention strategies (e.g., salary adjustments or training programs).

The solution included a SHAP-based explanation layer, helping HR understand the *why* behind every risk score (e.g., 'Lack of growth' vs 'Salary disparity'), making the AI's output actionable and fair.