tourism-package-model

Tourism Package Prediction Model

This model predicts whether a customer is likely to purchase the Wellness Tourism Package.

Model

• Algorithm: XGBoost binary classifier
• Preprocessing: numeric median imputation, standard scaling, categorical mode imputation, one-hot encoding
• Target: ProdTaken
• Decision use: prioritize sales outreach and campaign targeting

Training

Training uses stratified train/test splitting, class imbalance weighting, grid-search hyperparameter tuning, threshold analysis, and MLflow experiment tracking.

Evaluation

Local validation generated these representative test metrics:

• Accuracy: 0.913
• Precision: 0.764
• Recall: 0.794
• F1 score: 0.778

Limitations

The model should support human sales decisions rather than fully automate customer treatment. Predictions should be monitored over time because campaign strategy, package positioning, and customer behavior can change.