HR Data Analytics – Salary Prediction & Workforce Insights

A comprehensive data analytics and machine learning project exploring 5,000 employee records to understand salary drivers, workforce patterns, and predictive modeling for HR decision-making.

---

Dataset Overview

The dataset contains 5000 employee records with 20 features, including:

• Gender
• Business
• Dependancies
• Calls
• Contract Type
• Billing
• Rating
• Age
• Salary
• Base Pay
• Bonus
• Unit Price
• Volume
• Opening Balance
• Closing Balance
• Low
• Unit Sales
• Total Sales
• Months (Tenure)
• Education

Key Details

• Missing values found in Base Pay, Opening Balance, and Total Sales
• Handled using KNN Imputer
• Final cleaned dataset contains zero missing values

---

Data Preparation

Steps performed:

• Loaded CSV and verified data shape (5000 × 20)
• Identified columns with missing values
• Imputed missing values using KNN Imputer
• Applied Label Encoding to categorical features
• Checked variable distributions using boxplots
• Removed outliers using the IQR method
• Performed Spearman correlation for feature selection

---

Analysis & Visualizations

1. Gender Distribution

• Male: 2528
• Female: 2472
  Employee distribution is nearly equal.

2. Education Breakdown

• PG: 2979
• Graduation: 1980
• Intermediate: 27
• High School: 14

3. Contract Type Distribution

• Month-to-Month: 2777
• Two Year: 1195
• One Year: 1028

4. Salary Distribution & Outliers

• Boxplots used to inspect Salary, Base Pay, Bonus, etc.
• Significant outliers identified and removed.

5. Salary Correlations

Spearman correlation revealed:

• Total Sales → Salary: 0.99 (near perfect monotonic relation)
• Bonus, Base Pay, Unit Sales also highly correlated (> 0.8)

These became major predictors for modeling.

6. Numerical Feature Relationships

Scatter plots showed:

• Higher Total Sales → Higher Salary
• Higher Bonus → Higher Salary
• More Months of Experience → Higher Salary

7. Categorical Impact on Salary

• PG employees earn: ₹50,000 – ₹2,00,000
• High School / Intermediate: < ₹25,000
• Gender shows no meaningful salary difference.

8. Company Hiring Trends

• 765 hires in the last four months
• 269 employees have long tenure (72 months)

---

Machine Learning Models

Models trained after cleaning, encoding, scaling, and feature selection:

• Linear Regression: 78.27%
• Decision Tree Regression: 99.35%
• Random Forest Regression: 99.09%
• XGBoost Regression: 99.24% (best performer)

Cross-Validation Results

• Random Forest CV: 0.99993
• XGBoost CV: 0.99961

---

Final Model

• XGBoost Regressor selected as final model
• Normalized using StandardScaler
• Model saved as: hr.pkl
• Scaler saved as: schr.pkl
• Deployed using both Tkinter and Streamlit interfaces

---

Key Insights

Salary is most strongly influenced by:

• Age
• Education
• Months of Experience
• Total Sales
• Performance Metrics

Additional observations:

• Higher sales strongly increase salary
• Education level affects salary brackets significantly
• Gender has no major impact on salary
• Strong correlations enabled highly accurate ML predictions

---

Technologies Used

• Python
• Pandas
• NumPy
• Scikit-learn
• XGBoost
• Seaborn
• Matplotlib
• Pingouin
• Tkinter
• Streamlit

---

How to Run the Project

1. Clone the repository

bash git clone https://github.com/yourusername/HRDataAnalytics.git

2. Install dependencies

bash

Copy code

pip install pandas numpy matplotlib seaborn scikit-learn xgboost streamlit

3. Run the notebook

bash

Copy code

jupyter notebook HRDataAnalytics.ipynb

4. Launch the Streamlit App

bash

Copy code

streamlit run hrapp.py

Future Improvements

• Integrate SHAP for model explainability

• Add HR attrition prediction module

• Create dashboards using PowerBI or Streamlit

• Add advanced feature importance visualizations

• Deploy as a complete web-based prediction system

Contributing

Contributions are welcome.

Feel free to add new models, visualizations, or workflow improvements.

Support

If you found this project helpful, consider starring the repository.