nittygritty2106/travel_prediction_mlops

GradientBoosting Model for Travel Package Prediction

This model is the best performing model from an MLOps pipeline designed to predict customer purchases of a 'Wellness Tourism Package'.

Model Details

• Model Type: GradientBoosting
• Best ROC-AUC Score: 0.9725
• MLflow Run ID: a0015873740341eda992c37bb439ea61

Usage

This model can be loaded and used for inference on new customer data to predict their likelihood of purchasing the 'Wellness Tourism Package'.

Pipeline Overview

The MLOps pipeline involved:

1. Data Preprocessing: Handling missing values, encoding categorical features.
2. Model Training: Experimenting with various classical ML algorithms for tourism package prediction, including:
   • Decision Tree
   • Bagging
   • Random Forest
   • AdaBoost
   • Gradient Boosting
   • XGBoost
3. Model Selection: The GradientBoosting was selected based on its performance during Randomized Search Cross-Validation.[2]
4. MLflow Tracking: Logged parameters, metrics, and artifacts using MLflow for reproducibility and comparison.
5. Hugging Face Hub Deployment: The best model, along with its metadata and a README, is pushed to the Hugging Face Hub.

Metrics Captured

The following metrics were tracked for each model:

• Accuracy
• Precision
• Recall
• F1-score
• ROC-AUC