HR Employee Attrition and Workforce Analytics Dashboard

01 / PROBLEM

Employee attrition can affect workforce continuity and hiring costs. HR teams need ways to explore patterns in workforce data and identify areas that may warrant further investigation, without treating a model score as a decision about an individual.

02 / APPROACH

An interactive analytics dashboard built around the IBM HR Analytics Employee Attrition & Performance dataset. It combines exploratory analysis, visual comparisons, a baseline Logistic Regression model, and a filterable risk-analysis view.

03 / KEY FEATURES

• Interactive overview of workforce size and attrition patterns
• Demographic, work, compensation and tenure analysis
• Left-versus-stayed comparisons and exploration of associated factors
• Class-balanced Logistic Regression baseline with a stratified train/test split
• Model evaluation using accuracy, precision, recall and ROC-AUC
• Filterable risk-analysis list with CSV export
• Data-informed retention questions and recommendations

TECHNOLOGY

ARCHITECTURE

HR analytics dataset → cleaning and feature engineering → exploratory analysis → Logistic Regression baseline → model evaluation → interactive Streamlit dashboard and exportable risk-analysis view.

IMPLEMENTATION

The project prepares workforce features for analysis, explores attrition patterns through interactive visualizations, and evaluates a class-balanced Logistic Regression baseline using a stratified 75/25 train/test split. The dashboard organizes findings into focused views for workforce overview, demographics, compensation, attrition drivers, risk analysis and raw data.

OUTCOMES & NOTES

On the documented evaluation split, the baseline model achieved 78.0% accuracy, 64.4% recall, 38.8% precision and 0.81 ROC-AUC. These results describe this dataset and split only; they do not establish real-world predictive performance.

Uses the fictional IBM HR Analytics Employee Attrition & Performance dataset (1,470 employee records). Observed relationships do not establish causation. Risk scores should support review and conversation—not automated employment decisions or judgments about individual employees.