Advanced HR Analytics

Predicting employee attrition and optimizing workforce retention with AI-driven insights.

View Code on GitHub
SITUATION

The Challenge

High employee turnover was leading to increased recruitment costs and loss of institutional knowledge. HR teams lacked actionable data to identify at-risk employees and understand the underlying drivers of attrition.

TASK

The Objective

Develop a comprehensive analytics platform to visualize workforce demographics, analyze compensation trends, and build a predictive model to identify employees with a high risk of leaving the organization.

ACTION

The Solution

Built an interactive dashboard using R Shiny. Implemented a Random Forest model for attrition prediction. Designed intuitive visualizations with Plotly and ggplot2 to uncover insights into salary, satisfaction, and work-life balance.

RESULT

The Impact

Delivered a tool that provides real-time visibility into workforce health. The predictive model successfully identifies at-risk employees, enabling proactive retention strategies and data-driven decision-making for HR leadership.

Skills Demonstrated

R Programming

Data Manipulation & Logic

Shiny Framework

Interactive Web Apps

Predictive Modeling

Random Forest & Risk Scoring

Data Visualization

ggplot2 & Plotly

UI/UX Design

User-Centric Dashboarding

Key Insights

  • Overtime Impact: Employees working frequent overtime show a 30% higher attrition rate compared to peers.
  • Tenure Risk: The highest risk of turnover occurs within the first 2 years of employment, suggesting a need for better onboarding.
  • Role-Specific Turnover: Sales Representatives have the highest turnover, correlated with lower work-life balance scores.
  • Compensation: Below-market salary hikes are a strong predictor of attrition, even among satisfied employees.

Key Learnings

"One of the biggest takeaways was the importance of feature engineering. Creating a composite 'Risk Score' from multiple factors provided much more value to HR stakeholders than raw probabilities."

I also learned that dashboard performance is critical. Optimizing reactive expressions in Shiny significantly reduced load times for large datasets.

Live Dashboard Demo

Open in New Tab ↗

Built With

R Shiny Dashboard Plotly DT Random Forest CSS3