Skip to content

Latest commit

 

History

History
57 lines (43 loc) · 2.21 KB

README.md

File metadata and controls

57 lines (43 loc) · 2.21 KB

Salary Prediction Model

Project Overview

This project is a machine learning pipeline designed to predict employee salaries based on features like job title, department, years of experience, and other relevant factors. Utilizing powerful models like XGBoost and feature selection techniques such as Recursive Feature Elimination (RFE), the project aims to deliver accurate and insightful salary predictions.

Table of Contents

  • Features
  • Data
  • Model Evaluation
  • Contributing
  • License

Features

  • Data Preprocessing: Handling missing values, transforming skewed data, and scaling numeric data.
  • Exploratory Data Analysis (EDA): Analysis and visualization of key patterns in the dataset.
  • Feature Engineering and Selection: Using RFE to retain the most influential features for improved model performance.
  • Salary Prediction: Utilizing XGBoost and optimizing hyperparameters to produce reliable salary predictions.
  • Model Evaluation Metrics: Using Mean Squared Error (MSE) and Mean Absolute Error (MAE) to assess model accuracy.
  • Visualization: Graphical analysis of feature importance, distributions, and model predictions.

Data

The dataset was obtained from Kaggle and is licensed under the Apache 2.0 License on the Kaggle platform. It includes features such as:

Gender Department Overtime Hours Country Years of Experience Relationships between these factors and monthly salary information were explored using visual analysis, statistical methods, and AI-based techniques. As a result, a predictive model for salary estimation was developed.

Employee Dataset

Model Evaluation

The following metrics were used to assess model performance:

Mean Squared Error (MSE): 607,925.01 Mean Absolute Error (MAE): 678.04 Best Hyperparameters for XGBoost python

{
    'learning_rate': 0.01,
    'max_depth': 3,
    'n_estimators': 50,
    'subsample': 0.7
}

Contributing

Contributions are welcome! If you have suggestions for improvement or new features, please submit a pull request or open an issue.

License

This project is licensed under the Apache License 2.0. See the LICENSE file for more details.