Tourism Package Purchase Prediction

This project leverages data science to predict whether a customer is likely to purchase a travel package, empowering tourism businesses with actionable insights for marketing and product development.

Executive Summary

• Target: Predict binary outcome - Will the customer purchase the package?
• Dataset: 20 features including demographics, behavior, and preferences
• Goal: Boost marketing ROI by identifying potential buyers in advance

Business Insights

Exploratory data analysis revealed the following:

• Married customers are less likely to convert than single/unmarried
• Higher income and passport ownership strongly correlate with purchases
• Basic and standard packages showed the highest conversion rates
• Executives and younger customers tend to have higher conversion likelihood

Models & Performance

I evaluated Logistic Regression, SVM, Decision Tree, Random Forest, and SVD. Here are the final tuned metrics for my best-performing model:

Model	Precision	Recall	F1 Score	RMSE
SVM (RBF Kernel)	0.54	0.69	0.61	~
Logistic Regression	0.47	0.67	0.55	~
Random Forest	0.58	0.47	0.52	~
Decision Tree (Tuned)	0.52	0.62	0.57	~

Below is feature importance from the tuned Decision Tree (for interpretability) and the Random Forest (for more robust, aggregated insights). Both agree on key features: Monthly Income, Age, and Number of Trips.

Explore the Full Project

All code, data, and reports are available in the full GitHub repository.

How to Use

pip install -r requirements.txt
python scripts/model_pipeline.py