HR Analytics – Predicting Employee Attrition

📌 Objective

Use analytics to understand the main causes of employee resignation and predict future attrition.

---

1. Data Understanding & Preparation

• Imported and explored the HR dataset (1,470 employee records with 35 variables).
• Verified there were no missing values.
• Removed non-informative columns: EmployeeCount, EmployeeNumber, StandardHours, Over18.
• Converted the target variable Attrition into binary format (Yes = 1, No = 0).

---

2. Exploratory Data Analysis (EDA)

• Department-wise attrition: Highest in Research & Development, lowest in Human Resources.
• Salary bands: Higher attrition in lower salary band compared to medium and high.
• Promotions: Employees with fewer or no promotions in recent years were more likely to leave.
• Tenure: Shorter-tenure employees had higher attrition.
• Overtime: Employees working overtime were more prone to attrition.
• Marital Status: Single employees showed higher attrition than married or divorced employees.

---

3. Feature Engineering

• Created Income Bands → grouped into Low, Medium, High.
• Derived a Tenure feature → YearsAtCompany / TotalWorkingYears.
• Added a Promotion Gap → measured promotion delays for modeling.

---

4. Modeling

• Split dataset into training and testing (stratified by attrition).
• Preprocessing:
  • Standardized numerical variables.
  • One-hot encoded categorical variables.
• Built two models:
  • Logistic Regression → strong performance, identified key risk factors like Overtime, Low Salary, Job Role type.
  • Decision Tree → interpretable rules, key drivers included Total Working Years, Overtime, Daily Rate, Age.

---

5. Model Comparison

• Logistic Regression → better predictive accuracy and generalization.
• Decision Tree → slightly less accurate, but easier interpretability for HR decision-making.

---

6. Feature Importance (Decision Tree)

Top predictors of attrition:

• Total Working Years
• Overtime
• Monthly Income
• Daily Rate
• Age
• Categorical factors → Single employees, Job Role (Research Scientist)

---

7. Data Export for Visualization

📂 Data Sources

• Actual dataset → HR Employee Attrition.csv

• Transformed dataset → cleaned_hr_data.csv

  • Encoded + scaled features
  • Used for machine learning modeling

• Original dataset → hr_data_cleaned_original.csv

  • Retained categorical + numerical values
  • Used for Power BI dashboards and business-friendly reporting

---

8. Power BI Dashboard Insights

• Overall Workforce & Attrition: Majority are active, but attrition ~16%.
• Department-wise: R&D and Sales show highest attrition counts; HR shows proportionally high attrition despite smaller size.
• Income & Attrition: Lower-income employees leave most often; high attrition also seen among high-income Sales Executives.
• Job Role Trends: Laboratory Technicians, Research Scientists, Sales Reps → higher attrition at lower incomes; Sales Executives attrition across income levels.
• Career Stage: Younger employees most prone to leave (especially in Sales & HR); attrition decreases steadily with seniority.
• Other Key Drivers: Overtime and lack of promotions strongly linked with resignations.

---

🚀 Deliverables

• EDA insights on attrition by department, salary, promotions, overtime, demographics.
• Classification models (Logistic Regression & Decision Tree) with comparison.
• Feature importance analysis for HR policy recommendations.
• Power BI dashboard with attrition insights.
• Exported datasets for ML and visualization.