Enhanced accuracy
Fewer delays
Higher adoption
Streamlined operations
The challenge
Reforming carry-on predictions for effortless boarding
A major U.S. airline struggled with inaccurate carry-on baggage estimates, leading to boarding inefficiencies and delays. Their legacy model lacked real-time insights, limiting accuracy and adoption. To resolve this, they sought an advanced solution to optimize predictions, enhance gate agent decisions, and streamline the boarding process.
Key challenges
No real-time check-in or cabin space tracking
Rigid model hindered updates and fine-tuning
Agents avoided unclear, unreliable predictions
Errors reduced trust among gate agents
The solution
Smart carry-on prediction for hassle-free boarding
Enhanced predictive modeling
Improved forecasting with feature engineering
Built a CatBoost model for accuracy
Fixed under-prediction issues
Real-time insights
Integrated live data for dynamic updates
Shifted from batch to real-time processing
Provided range-based estimates for clarity
Implementation approach
1
Data and model development
Aggregated real-time and historical data
Enhanced prediction accuracy
Continuous model refinements
2
AWS-based deployment
Scalable, real-time processing
Hosted on AWS for efficiency
Optimized system for speed
3
User adoption
Boosted trust with precise insights
Intuitive front-end for agents
Reduced flight delays
The impact
Transforming boarding efficiency with AI-driven insights
Immediate benefits
Accurate bag predictions
Higher agent adoption
Fewer flight delays
Long-term benefits
Real-time insights
Efficient baggage handling
Better passenger experience
KPIs
Strong agent engagement
90% prediction accuracy
Smoother boarding
Looking ahead
Enhanced AI models
Continuous improvements for even higher accuracy
Expanded adoption
Broader rollout across more airports
Optimized operations
Further streamlining of boarding and baggage handling
