https://github.com/ayan6943/employee-attrition-prediction-with-machine-learning

Employee Attrition Prediction with Machine Learning | Analyzing HR data to predict employee turnover using Random Forest. Includes EDA, feature engineering, model training, and evaluation. Achieved 90% accuracy.
https://github.com/ayan6943/employee-attrition-prediction-with-machine-learning

attrition employee machine-learning matplotlib numpy pandas python randomforestclassifier scikit-learn seaborn smote

Last synced: 3 months ago
JSON representation

Employee Attrition Prediction with Machine Learning | Analyzing HR data to predict employee turnover using Random Forest. Includes EDA, feature engineering, model training, and evaluation. Achieved 90% accuracy.

• Host: GitHub
• URL: https://github.com/ayan6943/employee-attrition-prediction-with-machine-learning
• Owner: ayan6943
• Created: 2025-03-23T07:09:50.000Z (7 months ago)
• Default Branch: main
• Last Pushed: 2025-03-23T07:18:56.000Z (7 months ago)
• Last Synced: 2025-03-23T08:19:30.196Z (7 months ago)
• Topics: attrition, employee, machine-learning, matplotlib, numpy, pandas, python, randomforestclassifier, scikit-learn, seaborn, smote
• Language: Jupyter Notebook
• Homepage:
• Size: 242 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# 🧠 Employee Attrition Prediction with Machine Learning (Random Forest & XGBoost)

This project focuses on predicting employee attrition using two machine learning approaches — **Random Forest** and **XGBoost** — trained on the IBM HR Analytics dataset. It aims to help HR departments proactively identify at-risk employees and develop effective retention strategies.

By incorporating **SHAP explainability** with XGBoost, the project not only achieves high accuracy but also provides transparent insights into why employees may leave.

---

## 📊 Dataset

- **Source**: [IBM HR Analytics Employee Attrition Dataset](https://www.kaggle.com/datasets/pavansubhasht/ibm-hr-analytics-attrition-dataset)

- **Features**: Employee demographics, job role, income, overtime, satisfaction, etc.

- **Target**: `Attrition` (Yes/No)

---

## 🧰 Tools & Technologies

| Component       | Tools Used |

|------------------|-------------|

| Language         | Python |

| Libraries        | Pandas, NumPy, Matplotlib, Seaborn, Scikit-learn, XGBoost, SHAP |

| Models           | Random Forest, XGBoost |

| Explainability   | SHAP (for XGBoost) |

| Imbalance Handling | SMOTE |

---

## 🧪 Project Workflow

1. **EDA & Data Cleaning**

   - Visualized attrition patterns by role, overtime, satisfaction

   - Removed irrelevant/constant columns like `StandardHours`

   - Handled missing values and categorical variables

2. **Feature Engineering**

   - Identified key predictors

   - Encoded categorical variables and scaled numerical ones

3. **Modeling**

   - Trained both **Random Forest** and **XGBoost**

   - Addressed class imbalance using **SMOTE**

   - Evaluated using Accuracy, F1-Score, Precision, Recall, ROC-AUC

4. **Explainability (XGBoost only)**

   - Used **SHAP** to visualize feature importance and explain individual predictions

---

## 📈 Model Performance Comparison

| Metric        | Random Forest | XGBoost + SHAP |

|---------------|---------------|----------------|

| Accuracy      | 90%           | **92%**        |

| Precision     | 0.88 / 0.93   | **0.91 / 0.92** |

| Recall        | 0.93 / 0.87   | **0.93 / 0.91** |

| F1-Score      | 0.90          | **0.92**       |

| ROC-AUC       | 0.90          | **0.971** ✅    |

> 🚀 **XGBoost** outperformed Random Forest, especially in ROC-AUC and class balance, making it more reliable for real-world deployment.

---

## 🔍 SHAP Explainability (XGBoost)

- **SHAP Summary Plot**: Visualizes global feature importance

- Most influential features:

  - `OverTime`, `MonthlyIncome`, `JobSatisfaction`, `JobRole`

---

## 📄 License

This project is licensed under the MIT License.