Inference model for traveler classification

• US 10,282,797 B2
• Filed: 04/01/2015
• Issued: 05/07/2019
• Est. Priority Date: 04/01/2014
• Status: Active Grant

• Current Assignee: Amgine Technologies (US), Inc.
• Original Assignee: Amgine Technologies (US), Inc.
• Inventors: Valverde, Jr., L. James, Miller, Harold Roy, Miller, Jonathan David
• Primary Examiner(s): Casey, Alexis M
• Application Number:
  US14/676,302
• Publication Number:
  US 20150278970A1
• Time in Patent Office: 1,497 Days
• Field of Search: 705 267
• CPC Class Codes: G06Q 30/0631 Item recommendations
• CPC Class Codes: G06Q 50/14 Travel agencies

Abstract

A method for classifying a prospective traveler based on statistical inference is described herein. The method comprises receiving an input associated with the prospective traveler. Encoded representation of preferences and goals may be extracted from the input and levels and may be assigned to the preferences and goals. Based on the levels assigned to the preferences and goals, the prospective traveler may be classified according to one or more traveler profiles. Based on the classification, one or more consumer choices may be offered to the prospective traveler.

Claims

1. A computer-implemented method for classifying a prospective traveler based on a statistical inference, the method comprising:

receiving, by a processor, input associated with the prospective traveler, the input including at least free text data provided by the prospective traveler and data from one or more online resources associated with the prospective traveler, the one or more online resources being retrieved by a search engine;

parsing the received input by a parser to extract the free text data;

extracting, by the processor, an encoded representation of a plurality of preferences and goals from the free text data parsed by the parser and from the data from the one or more online resources retrieved by the search engine;

assigning, by the processor, one of a plurality of preference levels to each of the preferences and goals to define a preference structure of the prospective traveler, the preference structure represented by a vector comprising each of the assigned preference levels for each of the preferences and goals;

aggregating the assigned preference levels for each of the preferences and goals represented in the preference structure;

based on the aggregating the assigned preference levels for each of the preferences and goals represented in the preference structure;

determining a numerical value representing a highest probability of the prospective traveler fitting one or more predefined traveler profiles using one or more machine learning techniques; and

classifying the prospective traveler according to the one or more predefined traveler profiles;

developing an uncertain inference capability to classify the prospective traveler in terms of a first high level attribute of a plurality of high level attributes, wherein the uncertain inference capability comprises a probabilistic inference model representing conditional dependences between the first high level attribute and the preferences and goals associated with the first high level attribute;

constructing a multi-attribute inference model of consumer choice from the uncertain inference capability of the first high level attribute, an uncertain inference capability of a second high level attribute, and product characteristics scored;

determining a plurality of products with a high probability of being purchased by the prospective traveler using a triangulation statistical analysis of the product characteristics;

estimating a numerical probability value that the prospective traveler will choose one of the plurality of products, the numerical probability value estimated for at least two of the plurality of products; and

offering, by the processor, one or more consumer choices to the prospective traveler via graphical elements on a graphical user interface by displaying a predetermined number of the plurality of products to the prospective traveler on the graphical user interface, in order of the numerical probability value that the prospective traveler will choose the product, the processor being configured to receive a selection by the prospective traveler of the one or more offered consumer choices.

2. The method of claim 1, wherein the free text data includes one or more of the following:

natural language text, voice data, and oral exchange.

3. The method of claim 1, wherein the one or more predefined traveler profiles include at least a leisure traveler, a self-paying business traveler, a reimbursed traveler, and a business class traveler.

4. The method of claim 1, wherein the predefined traveler profile defines a choice of the prospective traveler according to time sensitivity, price sensitivity, and content affinity.

5. The method of claim 1, wherein the offering includes prioritizing the one or more consumer choices based on the one or more predefined traveler profiles.

6. The method of claim 1, further comprising:

receiving a response of the prospective traveler to the one or more consumer choices;

analyzing the response of the prospective traveler;

based on the analysis, determining that one or more purchase decisions of the prospective traveler contradicts the preference levels assigned to the preferences and goals;

based on the determining, re-assigning the preference levels to the preferences and goals; and

offering further one or more consumer choices to the prospective traveler.

16. The method of claim 1, wherein the plurality of high level attributes comprises at least two of price sensitivity, time sensitivity, and content affinity.

17. The method of claim 1, further comprising receiving a selection of the one or more consumer choices of the prospective traveler;

and reclassifying the prospective traveler based on the received selection.

18. The method of claim 1, wherein the one or more machine learning techniques comprise at least one of a neural network, SVM, Bayesian, and naï

ve Bayesian.

19. The method of claim 1, wherein the plurality of preferences and goals include at least one of:

a class of service requested, a booking delta, a ticket fare level, a connection count, a travel speed, a connection time sum, and a departure time.

7. A system for classifying a prospective traveler based on statistical inference, the system comprising:

a processor configured to;

receive input associated with the prospective traveler, the input including at least free text data provided by the prospective traveler and data from one or more online resources associated with the prospective traveler, the free text data being parsed by a parser and the data from the one or more online resources being retrieved by a search engine;

extract encoded representation of preferences and goals from the free text data parsed by the parser and the data from the one or more online resources retrieved by the search engine;

assign levels to the encoded representation of preferences and goals to define a preference structure of the prospective traveler;

classify the prospective traveler according to one or more traveler profiles based on the levels assigned to the encoded representation of preferences and goals by matching the preference structure of the prospective traveler to one or more preference structures associated with the one or more traveler profiles, the classifying performed with at least one or more machine learning techniques;

develop an uncertain inference capability to classify the prospective traveler in terms of a first high level attribute of a plurality of high level attributes, wherein the uncertain inference capability comprises a probabilistic inference model representing conditional dependences between the first high level attribute and the preferences and goals associated with the first high level attribute;

construct a multi-attribute inference model of consumer choice from the uncertain inference capability of the first high level attribute, an uncertain inference capability of a second high level attribute, and product characteristics scored, the multi-attribute inference model being based on computations for consumer choice dimensions, the multi-attribute inference model being applied to identify one or more consumer choices to be offered to the prospective traveler;

determine a plurality of products with a high probability of being purchased by the prospective traveler using a triangulation statistical analysis of the product characteristics;

estimate a numerical probability value that the prospective traveler will choose one of the plurality of products, the numerical probability value estimated for at least two of the plurality of products; and

offer the one or more consumer choices to the prospective traveler via graphical elements on a graphical user interface by displaying a predetermined number of the plurality of products to the prospective traveler on the graphical user interface, in order of the numerical probability value, the processor being configured to receive a selection by the prospective traveler of the one or more offered consumer choices; and

a database in communication with the processor configured to store at least the input and the preferences and goals.

8. The system of claim 7, wherein the free text data includes one or more of the following:

natural language text, voice data, and oral exchange.

9. The system of claim 7, wherein the one or more traveler profiles include at least a leisure traveler, a self-paying business traveler, a reimbursed traveler, and a business class traveler.

10. The system of claim 7, wherein the one or more traveler profiles define a choice of the prospective traveler according to time sensitivity, price sensitivity, and content affinity.

11. The system of claim 10, wherein the encoded representation of preferences and goals are associated with the time sensitivity, wherein the encoded representation of preferences and goals associated with the time sensitivity include one or more of the following:

a class of service requested, a booking delta, a ticket fare level, a connection count, a travel speed, a connection time sum, and a departure time.

12. The system of claim 10, wherein the encoded representation of preferences and goals are associated with the price sensitivity, wherein the encoded representation of preferences and goals associated with the price sensitivity include one or more of the following:

a class of service requested, a booking delta, a ticket fare level, a connection count, a travel speed, a connection time sum, and a departure time.

13. The system of claim 10, wherein the encoded representation of preferences and goals are associated with the content affinity, wherein the encoded representation of preferences and goals associated with the content affinity include one or more of the following:

a class of service requested and a ticket fare level.

14. The system of claim 10, wherein the multi-attribute inference model is based on a multi-attribute value function.

15. A system for classifying a prospective traveler based on statistical inference, the system comprising:

a processor configured to;

receive input associated with the prospective traveler, the input including at least free text data provided by the prospective traveler and data from one or more online resources associated with the prospective traveler, the free text data being parsed by a parser, the data from the one or more online resources being retrieved by a search engine;

extract encoded representation of preferences and goals from the free text data parsed by the parser and the data from the one or more online resources retrieved by the search engine;

assign levels to the encoded representation of preferences and goals to identify a preference structure of the prospective traveler;

classify the prospective traveler according to one or more traveler profiles based on the levels assigned to the encoded representation of preferences and goals by matching the preference structure of the prospective traveler to one or more preference structures associated with the one or more traveler profiles, the classifying performed with at least one or more machine learning techniques;

develop an uncertain inference capability to classify the prospective traveler in terms of a first high level attribute of a plurality of high level attributes, wherein the uncertain inference capability comprises a probabilistic inference model representing conditional dependences between the first high level attribute and the preferences and goals associated with the first high level attribute;

construct a multi-attribute inference model of consumer choice from the uncertain inference capability of the first high level attribute, an uncertain inference capability of a second high level attribute, and product characteristics scored;

determine a plurality of products with a high probability of being purchased by the prospective traveler using a triangulation statistical analysis of the product characteristics;

estimate a numerical probability value that the prospective traveler will choose one of the plurality of products, the numerical probability value estimated for at least two of the plurality of products;

identify one or more consumer choices to the prospective traveler using the inference models; and

offer the one or more consumer choices to the prospective traveler via graphical elements on a graphical user interface by displaying a predetermined number of the plurality of products to the prospective traveler on the graphical user interface, in order of the numerical probability value, the processor being configured to receive a selection by the prospective traveler of the one or more offered consumer choices; and

a database in communication with the processor configured to store at least the input, the preferences and goals, and the inference models.