README.md

metadata

title: EmployeeAttritionClassification
emoji: 🏢
colorFrom: indigo
colorTo: blue
sdk: docker
app_port: 8501
tags:
  - streamlit
pinned: false
short_description: Employee Attrition Predictor using XGBoost.
license: mit

Overview

This app predicts the probability of employee attrition (Attrition = 1) using a trained XGBoost classifier. The model was evaluated using ROC-AUC and achieved strong validation performance.

Files in this repo

Place these files in the repo root (same folder as app.py):

• xgb_model.pkl — trained XGBoost model
• feature_names.pkl — list of training feature columns (order matters)
• threshold.pkl — decision threshold (e.g., 0.35)

How to use the app

1) Single prediction (form)

• Enter numeric/ordinal feature values
• Select one option for categorical groups (one-hot)
• The app outputs:
  • Attrition probability
  • Attrition prediction using your saved threshold

2) Batch prediction (CSV upload)

Upload a CSV that is already in the same feature format as your training data (after preprocessing and one-hot encoding). The app will align columns automatically using feature_names.pkl.

Notes

• Kaggle submissions typically require probabilities, not class labels.
• If you want to change the decision threshold, update threshold.pkl.

Local run

pip install -r requirements.txt
streamlit run app.py