🕵️‍♂️ HR Employee Attrition Prediction & Analysis

📌 Project Overview

Employee attrition is a critical challenge for organizations, leading to high replacement costs and loss of valuable talent. This project aims to build a robust Machine Learning pipeline to predict which employees are at risk of leaving, allowing the HR department to take proactive retention measures.

🚀 Business Value & Strategy

Instead of optimizing for overall accuracy (which is misleading in highly imbalanced HR datasets), this project prioritizes Recall. By implementing a Linear SVM paired with SMOTEENN, the final model successfully identifies ~81% of at-risk employees, providing HR with a highly actionable "watch list" to conduct stay-interviews before resignation letters are drafted.

🛠️ Tech Stack & Methodologies

• Language: Python
• Libraries: Pandas, Scikit-learn, Imbalanced-learn, Matplotlib, Seaborn
• Data Preprocessing: Custom Pipelines, RobustScaler, OneHotEncoding.
• Handling Imbalanced Data: SMOTEENN (Synthetic Minority Over-sampling Technique + Edited Nearest Neighbours).
• Modeling: Logistic Regression, Random Forest, KNN, and Support Vector Machines (SVM).
• Hyperparameter Tuning: GridSearchCV with Stratified K-Fold Cross Validation.

📊 Key Findings (Feature Importance)

By extracting coefficients from the Tuned Linear SVM, we identified the top drivers of employee turnover:

1. Job Role (Laboratory Technician): Highest risk factor, indicating potential issues with compensation or role design.
2. OverTime: Both excessive overtime and strict "no overtime" policies correlate with higher attrition (burnout vs. disengagement).
3. Business Travel: Frequent travelers and non-travelers leave more often than those who travel moderately.
4. Role Stagnation (YearsInCurrentRole): Lack of internal mobility significantly drives talent away.

📈 Model Performance

• Algorithm: Linear Support Vector Machine (SVM)
• Threshold: Default (0.50) - chosen for best practical HR balance.
• Recall (Class 1): 0.809 (Caught 29 out of 47 leaving employees)
• ROC AUC: 0.830

📂 Repository Structure

• V11_Employee_Attrition_Performance_prediction.ipynb : The complete End-to-End code (EDA, Pipeline, Modeling, Evaluation).
• WA_Fn-UseC_-HR-Employee-Attrition.csv : The HR dataset used for training and testing. Kaggle

💡 How to Run

1. Clone the repository: git clone https://github.com/omarDlgaber/HR-Attrition-Prediction.git
2. Install the required dependencies.
3. Run the Jupyter Notebook.