top of page
Search

🧠 Flight Risk Employee Prediction Project

  • Dr Dilek Celik
  • 5 days ago
  • 2 min read

In collaboration with Retainable, I developed a machine learning-based solution to help enterprise clients identify top talent at risk of leaving. The core idea was to support HR and leadership teams with a predictive decision-support tool that helps retain key employees before it’s too late.

While proprietary client data was used in production, this public-facing demo leverages a Kaggle HR dataset to simulate real-world deployment and outcomes.


In today’s competitive talent market, retaining top performers is more critical than ever. This project tackles that challenge head-on by building a predictive model that estimates the probability of an employee leaving, helping HR teams take proactive retention actions.


🔍 Project Summary

This end-to-end data science project delivers a fully interactive Streamlit app that predicts employee churn using real HR data. The model is powered by an XGBoost classifier, fine-tuned with Bayesian Optimization for optimal performance.

The app enables users to:

  • Input employee metrics like satisfaction level, average monthly hours, and years at the company

  • Instantly receive the predicted churn probability

  • Make data-informed decisions about employee retention risk


🧠 Key Features

  • Model: XGBoost Classifier with hyperparameter tuning via Bayesian Optimization

  • Performance: Achieved over 99% accuracy in validation

  • Explainability: Integrated SHAP (SHapley Additive exPlanations) to visualize feature importance

  • Deployment: Deployed via Streamlit Cloud for public access

  • User Interface: Intuitive layout allowing HR managers or business users to interact with the model in real time


🛠️ Technologies Used

  • Python

  • scikit-learn

  • XGBoost

  • SHAP

  • Streamlit

  • Git/GitHub


💡 Why It Matters

At Retainable, this prototype laid the groundwork for a broader flight risk prediction system, helping clients take data-driven actions to retain high performers. This demonstration version shows how even public data can drive meaningful insights and showcase solution value.

Simulation of Flight Risk Employee Prediction Application Using XGBoost Model
Simulation of Flight Risk Employee Prediction Application Using XGBoost Model
Use of Shap to analyse feature importance.
Use of Shap to analyse feature importance.
Feature importance with values
Feature importance with values

📂 Codebase (optional): GitHub Repository 


Comments

bottom of page