Travel-related prediction system

• US 8,200,514 B1
• Filed: 02/15/2007
• Issued: 06/12/2012
• Est. Priority Date: 02/17/2006
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Farecast, Inc. (Microsoft Corporation)
• Inventors: Crean, Hugh, Bartot, Jay, Fridgen, Michael, Hsu, David
• Primary Examiner(s): Hayes, John
• Assistant Examiner(s): EPSTEIN, BRIAN M
• Application Number:
  US11/675,499
• Time in Patent Office: 1,944 Days
• Field of Search: 705/5, 705/6, 705/27, 705/10, 705/26, 707/1
• US Class Current: 705/5
• CPC Class Codes: G06Q 10/02 Reservations, e.g. for tickets, services or events
• CPC Class Codes: G06Q 10/025 Coordination of plural reservations, e.g. plural trip segments, transportation combined with accommodation
• CPC Class Codes: G06Q 30/0202 Market predictions or forecasting for commercial activities
• CPC Class Codes: G06Q 30/0206 Price or cost determination based on market factors
• CPC Class Codes: G06Q 30/0629 for generating comparisons

Abstract

A method and system is provided for making price-related predictions of items based on training a classifier using historical information including pricing information for the items and labels relating to the price-related information that are generated automatically from the historical information. A prediction system provides price-related predictions for airline tickets. The prediction system collects flight information for flights on a daily basis. The prediction system generates training data to train a classifier to make the price-related predictions using the flight information. After training the classifier, the prediction system then can make the price-related predictions for a trip.

Claims

1. A computing system for making a price-related prediction of airline ticket prices for trips, comprising:

a memory storing computer-executable instructions of;

a component that provides flight information for airline flights for trips on various observation dates, each trip having one or more available flights, each trip having a departure airport and date and a return airport and date, wherein an observation date for flight information of a flight represents the date on which the flight information was current, the provided flight information for a flight including an actual price of the flight as of each observation date;

a component that generates a feature vector for each trip for each observation date based on the provided flight information;

a component that determines for each observation date for a trip what the price-related prediction should have been for the trip on each observation date by analyzing the actual prices of the flights of the trip as of later observation dates so that the determined price-related prediction accurately factors in the actual prices of the trip as of the later observation dates;

a component that trains a classifier to make a price-related prediction for trips using the feature vectors and determined predictions for the prices as of observation dates for the trips; and

a component that generates a price-related prediction for a target trip by applying the trained classifier to a feature vector generated based on flight information of the target trip; and

a processor for executing the computer-executable instructions stored in the memory.

2. The computing system of claim 1 wherein the price-related prediction is direction of airline ticket price.

3. The computing system of claim 2 wherein the component that trains the classifier trains a separate classifier for each possible direction of airline ticket price.

4. The computing system of claim 1 wherein the price-related prediction is whether price protection should be offered.

5. The computing system of claim 1 wherein the price-related prediction is whether a future-lower priced fare should be offered.

6. The computing system of claim 1 wherein the component that trains a classifier trains a separate classifier for each market.

7. The computing system of claim 1 wherein the component that generates feature vectors generates core features and derived features.

8. The computing system of claim 7 wherein the core features of a trip are derived from flight information of flights of the trip having the same observation date.

9. The computing system of claim 8 wherein the core features of a trip are selected from the group consisting of market, departure date, return date, minimum airline ticket price, average airline ticket price, maximum airline ticket price, number of flights with an airline ticket price near the minimum ticket price, and total number of available seats on all flights with an airline ticket price near the minimum ticket price.

10. The computing system of claim 7 wherein the derived features of a trip are derived from flight information of flights of the trip for multiple observation dates.

11. The computing system of claim 10 wherein the derived features of a trip are selected from the group consisting of moving averages of the core features, differences between derived features and core features, and differences between core features of the trip of the observation date and core features of the trip of prior observation dates.

12. The computing system of claim 1 including a component that receives from a user an identification of a target trip, retrieves a generated price-related prediction for the target trip, and presents to the user an indication of the retrieved price-related prediction.

13. The computing system of claim 1 wherein the price-related predictions for trips for an observation date are generated in batch mode.

14. The computing system of claim 13 wherein the price-related predictions for a trip are generated assuming different current airline ticket prices for the trip.

15. A computing system for predicting direction of airline ticket prices for trips, comprising:

a memory storing computer-executable instructions of;

a component that provides flight information for airline flights for trips on various observation dates, each trip having one or more available flights;

a component that generates a feature vector for each trip for each observation date based on the provided flight information;

a component that assigns a label to each trip for each observation date with enough subsequent observation dates to generate a label indicating what the price direction prediction should have been for the trip on the observation date, the label for each trip for each observation date being derived from the provided flight information for that trip from subsequent observation dates; and

a component that trains a classifier to make price direction predictions for trips using the feature vectors and their labels as training data; and

a processor for executing the computer-executable instructions stored in the memory.

16. A computing system for making price-related predictions for airline trips, comprising:

a memory storing computer-executable instructions of;

a component that provides flight information for airline flights for trips on various observation dates, a trip having one or more available flights;

a component that generates trip information for a trip by aggregating the flight information for the airline flights associated with the trip, the trip information being represented as feature vectors;

a component that assigns a label to each trip for each observation date with enough subsequent observation dates to generate a label indicating what the price-related prediction should have been for the trip on the observation date, the label for each trip for each observation date being derived from the provided flight information for that trip from subsequent observation dates;

a component that trains a classifier to make price-related predictions for trips using the feature vectors and their labels as training data; and

a component that generates a price-related prediction for a trip using a feature vector for the trip and the trained classifier; and

a processor for executing the computer-executable instructions stored in the memory.

17. The computing system of claim 16 wherein a separate classifier is trained for each market, a market being a combination of a departure location and a return location.

18. The computing system of claim 16 wherein a single classifier is trained for multiple markets, a market being a combination of a departure location and a return location.

19. The computing system of claim 16 wherein a classifier is trained for a market using feature vectors and labels of a different market, a market being a combination of a departure location and a return location.