📊 Employee Attrition Prediction & Analytics Dashboard - IBM HR Analytics

This project analyzes the IBM HR Analytics Employee Attrition & Performance dataset and builds multiple machine learning models to predict whether an employee is likely to leave the company. It includes data preprocessing, outlier handling, feature engineering, scaling, ML model comparison, and a Tableau dashboard for visualization.

---

📁 Dataset

• Source: Kaggle – IBM HR Analytics Employee Attrition
• Rows: 1470
• Columns: 35
• Target Variable: Attrition (Yes/No)

---

🛠️ Project Workflow

1. 🔍 Data Exploration

• Inspected data types, null values, summary statistics
• Removed constant columns (Over18, EmployeeCount, StandardHours)
• Visualized distributions using boxplots and summary charts

2. ✂️ Outlier Detection & Removal

Applied IQR (Interquartile Range) method to detect and remove outliers across all numeric columns, resulting in a clean dataset suitable for modeling.

3. 🏷️ Label Encoding

Converted categorical attributes into numeric codes:

✔️ Binary Columns

Encoded using LabelEncoder Examples:

• Attrition
• Gender
• OverTime

✔️ Multi-class Columns

Fields like BusinessTravel, JobRole, Department, etc., mapped to numerical values.

4. 📏 Feature Scaling

Used Min-Max Scaling to normalize all numeric columns into a 0–1 range. Exported final dataset as df_min_max.csv.

5. 📊 Tableau Dashboard

Created interactive charts and dashboards to visualize:

• Employee demographics
• Gender balance
• Department vs role distributions
• Monthly income patterns
• Attrition insights
• Heatmaps for multi-category patterns

Link to the Dashboard

6. 🤖 Machine Learning Models

🔧 Feature Selection

Selected top-correlated features (>0.1 correlation with target).

🧪 Train/Test Split

80% training, 20% testing.

🧠 Models Implemented

• Logistic Regression
• Random Forest
• SVM
• Decision Tree
• KNN

Each evaluated using:

• Accuracy
• Precision
• Recall
• F1-Score
• Confusion Matrix

Compared all models visually via bar chart.

---

📈 Results Summary

• Clear model comparison
• Visual confusion matrices
• Random Forest generally performs strongest
• Identified dominant attrition factors such as overtime, satisfaction levels, income, and tenure

---

🏁 Conclusion

This project delivers:

• A complete preprocessing and ML pipeline
• Feature encoding/scaling automation
• Multi-model evaluation
• HR analytics dashboard
• Practical insights to reduce employee attrition

---

🚀 How to Run

1. Clone repo
2. Install dependencies
3. Run notebooks
4. View Tableau dashboard

---