Employee Attrition Analytics

An end-to-end machine learning project that analyzes employee attrition using a dataset of 20,000 employee records. The project covers exploratory data analysis, classification modeling to predict attrition, K-Means clustering to segment the workforce, and regression modeling to understand income drivers.

Project Overview

Employee attrition is one of the costliest problems an HR team deals with — replacing an employee can cost anywhere from 50% to 200% of their annual salary once recruiting, onboarding, and lost productivity are accounted for. This project builds a complete analytics pipeline to help answer three questions:

• Who is likely to leave? — solved with classification models (Logistic Regression, Decision Tree, Random Forest)
• What natural groups exist in the workforce, and which group is riskiest? — solved with K-Means clustering
• What drives an employee's income? — solved with regression models (Linear Regression, Random Forest Regressor)

Repository Structure

employee-attrition-analytics/
│
├── Datasets/
│   ├── capstone_employee_attrition_dataset.csv    # Raw dataset (20,000 records, 29 columns)
│   └── capstone_data_dictionary.csv               # Column definitions and data types
│
├── Docs/
│   ├── Capstone_Problem_Statement_and_Data_Dictionary.pdf   # Original problem statement
│   └── Project_Workflow_Report.pdf                          # Concept-level write-up of the full workflow
│
├── Output_Plots/
│   ├── attrition_overview.png          # Overall & department-wise attrition rate
│   ├── correlation_matrix.png          # Correlation heatmap of numeric features
│   ├── income_overtime_attrition.png   # Income & overtime vs attrition
│   ├── confusion_matrix_lr.png         # Confusion matrix — Logistic Regression
│   ├── roc_curves.png                  # ROC curves — all classification models
│   ├── feature_importance.png          # Top features driving attrition (Random Forest)
│   ├── elbow_curve.png                 # Elbow method for choosing k
│   ├── silhouette_scores.png           # Silhouette score vs k
│   └── cluster_attrition.png           # Attrition rate by cluster
│
├── employee-attrition-analytics.py     # Full project code (EDA → preprocessing → modeling)
└── README.md

Dataset

• Source file: capstone_employee_attrition_dataset.csv
• Size: 20,000 employee records, 29 columns
• Target variable: attrition (Yes / No)
• Data dictionary: see Datasets/capstone_data_dictionary.csv for full column definitions

Workflow

1. Exploratory Data Analysis — overall and department-wise attrition rate, correlation analysis, income/overtime vs attrition
2. Preprocessing — feature/target split, scaling (StandardScaler), encoding (OneHotEncoder), stratified train-test split
3. Classification — Logistic Regression, Decision Tree, and Random Forest, evaluated on precision, recall, F1-score, and ROC-AUC; hyperparameter tuning via GridSearchCV
4. Model Interpretation — confusion matrix, ROC curve comparison, Random Forest feature importance
5. Clustering — K-Means with elbow method and silhouette score to choose the optimal number of clusters, followed by attrition-rate-by-cluster analysis
6. Regression — Linear Regression and Random Forest Regressor to predict monthly income and identify its strongest drivers

A full concept-level explanation of each step and why it was done is available in Docs/Project_Workflow_Report.pdf.

Tech Stack

• Language: Python
• Libraries: pandas, NumPy, scikit-learn, matplotlib, seaborn
• Environment: Google Colab

How to Run

1. Clone the repository

   git clone https://github.com/<your-username>/employee-attrition-analytics.git
   

2. Open employee-attrition-analytics.py in Google Colab or your local Python environment
3. Make sure capstone_employee_attrition_dataset.csv (from the Datasets/ folder) is available in the working directory
4. Run the script/notebook cells from top to bottom

Key Outputs

All generated charts are available in the Output_Plots/ folder, including attrition breakdowns, the correlation heatmap, confusion matrix, ROC curves, feature importance ranking, and clustering diagnostics (elbow curve, silhouette scores, cluster-wise attrition).

Author

Swaroop Kumar Vathada Data Analyst | BI Developer

🔗 LinkedIn