Overview
An automated travel itinerary system that combines K-means clustering and Random Forest algorithms to generate complete travel plans including destinations, hotels, and activities. The system processes client preferences and automatically produces editable PDF itineraries, reducing planning time from hours to minutes.
The Problem
Travel agencies spend more that a hour manually creating each itinerary, leading to high operational costs and inconsistent quality. The client needed a solution that could automate this process while maintaining flexibility for customizations.
Technical Solution
System Architecture
- •Two-phase clustering pipeline (destinations → hotels)
- •Random Forest classification for hotel recommendations
- •PHP backend for PDF generation and admin interface
- •Elbow method + silhouette scoring for optimal clusters
ML Framework
Scikit-learn
Data Points
1000+ historical itineraries
Processing Time
<1 minute per itinerary
Output Format
Editable PDF
Results & Impact
85%
Time savings per itinerary
< 1
Minute generation time
6
Destination clusters
Lessons Learned
Hybrid Approach Wins
Combining clustering (K-means) with classification (Random Forest) outperformed pure clustering solutions.
Minimum Viable Data
Achieved 85% accuracy with limited data points and it can massively be improved with more historical itineraries.
PHP Integration
Wrapping Python models in PHP reduced deployment costs by 60% compared to pure Python solutions