Flight Choice Prediction

API version: 2.0.2

Auth

No authentication required.

Base URL

https://test.api.amadeus.com/v2

Setup

1. No auth setup needed
2. POST /shopping/flight-offers/prediction -- create first prediction

Endpoints

1 endpoints across 1 groups. See references/api-spec.lap for full details.

shopping

Method	Path	Description
POST	/shopping/flight-offers/prediction	Predict the choice of flight offers.

Enhanced Skill Content

Question Mapping

• "Which flight offer is most likely to be booked?" -> POST /shopping/flight-offers/prediction
• "Can you rank these flight offers by booking probability?" -> POST /shopping/flight-offers/prediction
• "What is the choice prediction score for my flight search results?" -> POST /shopping/flight-offers/prediction
• "How likely is a traveler to pick this flight?" -> POST /shopping/flight-offers/prediction
• "Which of these itineraries has the highest conversion rate?" -> POST /shopping/flight-offers/prediction
• "Sort flight offers by predicted traveler preference" -> POST /shopping/flight-offers/prediction
• "Add machine learning predictions to my flight search results" -> POST /shopping/flight-offers/prediction
• "Which cheap flight is actually worth recommending?" -> POST /shopping/flight-offers/prediction
• "Enrich flight offers with choice probability scores" -> POST /shopping/flight-offers/prediction
• "I have flight offers from a search -- which ones should I show first?" -> POST /shopping/flight-offers/prediction
• "What does the prediction model say about overnight vs direct flights?" -> POST /shopping/flight-offers/prediction
• "How do I filter flight offers to only high-probability choices?" -> POST /shopping/flight-offers/prediction (then filter client-side by choiceProbability)

Response Tips

• Prediction results: Each flight offer in the response is enriched with a choiceProbability field (0.0-1.0). Higher means more likely to be booked. The original offer data is preserved alongside the prediction score.
• Errors (400): Malformed request body or missing required flight offer fields. Check that the body matches the Flight Offers Search response schema -- the prediction endpoint expects the same structure returned by GET /shopping/flight-offers.
• Errors (500): Amadeus server-side failure. Retry with exponential backoff; do not modify the request body.
• No pagination: This endpoint returns all enriched offers in a single response. Result count matches the number of offers submitted.

Anomaly Flags

• Low prediction spread: If all offers return nearly identical choiceProbability values (variance < 0.05), the model has low confidence -- surface this to the user as "predictions are inconclusive for this route/date."
• Rate limit headers: Watch for X-RateLimit-Remaining dropping below 10. Proactively warn before hitting the Amadeus free-tier cap (especially on test environment).
• Empty or truncated offers: If the response contains fewer offers than submitted, flag missing entries -- the model may have silently dropped malformed offers.
• Stale input data: If flight offers passed to prediction were fetched more than 15 minutes ago, prices and availability may have shifted. Flag that predictions may not reflect current inventory.
• HTTP 500 spikes: If repeated 500s occur, the prediction model may be temporarily unavailable. Surface this rather than silently retrying indefinitely.

Playbook

1. Search Flights Then Rank by Prediction

1. Call GET /v2/shopping/flight-offers with origin, destination, date, and traveler count
2. Take the full JSON response body (the data array of flight offers)
3. Pass it as the request body to POST /v2/shopping/flight-offers/prediction
4. Sort returned offers by choiceProbability descending
5. Present the top 3-5 results to the user

2. Filter to High-Confidence Recommendations Only

1. Obtain flight offers (via search or cache)
2. Submit to POST /v2/shopping/flight-offers/prediction
3. Filter results to only those with choiceProbability >= 0.7
4. If fewer than 2 offers pass the threshold, lower it to 0.5 and re-filter
5. Present filtered results with the prediction score shown as a percentage

3. Compare Direct vs Connecting Flight Preference

1. Search flight offers ensuring results include both direct and 1-stop itineraries
2. Submit all offers to POST /v2/shopping/flight-offers/prediction
3. Group results by number of stops (segments count minus one)
4. Compare average choiceProbability across groups
5. Report which flight type the model predicts travelers prefer for this route

4. Batch Prediction for Multiple Routes

1. For each origin-destination pair, call the flight offers search endpoint
2. Collect all offer sets
3. Submit each set separately to POST /v2/shopping/flight-offers/prediction (one call per route -- the endpoint does not support mixed routes)
4. Aggregate top-scoring offers across all routes into a single ranked list
5. Present cross-route recommendations sorted by prediction score

Response Tips

• Check response schemas in references/api-spec.lap for field details
• Create/update endpoints typically return the created/updated object

References

• Full spec: See references/api-spec.lap for complete endpoint details, parameter tables, and response schemas

Generated from the official API spec by LAP