This application offers two distinct machine learning approaches for predicting inbound shipments:
đ¯ XGBoost (Relationship-Based)
What it does: Analyzes the relationship between current shipments in your system and expected inbound shipments.
Best for: Operational decisions, real-time insights, tactical planning
đ SARIMA (Time Series)
What it does: Analyzes historical patterns to forecast future shipment volumes based on trends and seasonal patterns.
Best for: Strategic planning, seasonal analysis, capacity management
đ When to Use Each Model
Scenario
Recommended Model
Why
"How many inbounds if we have X shipments at noon?"
XGBoost
Relationship-based, real-time insights
"What's our expected volume next week?"
SARIMA
Captures seasonal patterns and trends
"Planning staffing for next month"
SARIMA
Long-term forecasting with seasonality
"Should we adjust operations based on current load?"
XGBoost
Immediate operational decisions
"Understanding weekly volume patterns"
SARIMA
Time series analysis of patterns
"Predicting based on current system state"
XGBoost
Feature-based relationship modeling
đĄ Pro Tip: Both models complement each other and provide different perspectives on your shipment data. XGBoost excels at understanding immediate operational relationships, while SARIMA captures longer-term patterns and seasonal trends.
â ī¸
Error Details
Stack Trace:
Predictive Tools
+ Advanced Options
XGBoost: Predicts inbound shipments based on the relationship between current shipments in system and historical inbound patterns. Best for understanding operational relationships.
Check this option to train a new model even if a cached model exists.
Where did each client's consumed guesses come from?
Cross-Client Flows
All guess transfers between different clients
Donor
â
Recipient
Count
Sprinkle Errors by Client
Shipments that could not be dynamically added to routes ("no-routes")
Errors by Reason
Why shipments could not be sprinkled onto routes
Donation vs. No-Route Correlation
Identifying clients who may have suffered no-routes due to donating guess spots
No-Routes per Donation
For net donors: how many no-routes occurred per guess spot donated away
How to interpret: A high "no-routes per donation" ratio suggests a client may have experienced more sprinkle failures
because their guess spots were consumed by other clients. This metric is only calculated for net donors (clients who donated more
spots than they received).
Sprinkle Geographic View
Visualize sprinkle successes and failures on routes
Predictive Tools