TRAJECTORY MODELING FOR CONTEXTUAL RECOMMENDATION

• US 20200132485A1
• Filed: 10/26/2018
• Published: 04/30/2020
• Est. Priority Date: 10/26/2018
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Zhao, Shiwan, Wang, Zhi Hu, Sun, Changhua, Su, Zhong
• Granted Patent: US 10,876,850 B2
• Time in Patent Office: Days
• CPC Class Codes: G01C 21/3476 using point of interest [POI] information, e.g. a route passing visible POIs
• CPC Class Codes: G01C 21/3484 Personalized, e.g. from learned user behaviour or user-defined profiles
• CPC Class Codes: G01C 21/3679 Retrieval, searching and output of POI information, e.g. hotels, restaurants, shops, filling stations, parking facilities G01C21/3611 takes precedence
• CPC Class Codes: G06N 3/044 Recurrent networks, e.g. Hopfield networks
• CPC Class Codes: G06N 3/084 Backpropagation, e.g. using gradient descent

Abstract

A computer-implemented method, computer program product, and computer processing system are provided for computing a trajectory-based Point of Interest recommendation. The method includes generating, by a processor device, a set of embeddings. Each of the embeddings in the set relates to a respective different trajectory contextual element of a user trajectory. The method further includes computing, by the processor device based on the set of embeddings, an activity representation that includes a set of POI candidate embeddings. The method also includes composing, by the processor device, a stop embedding based on the activity representation and the embeddings in the set and corresponding to a given stop in the user trajectory. The method additionally includes computing, by the processor device, the trajectory-based POI recommendation using an attention-based, user-specific, multi-stop trajectory, Recurrent Neural Network (RNN) model applied to the stop embedding.

Claims

1. A computer-implemented method for computing a trajectory-based Point of Interest recommendation, the method comprising:

generating, by a processor device, a set of embeddings, each of the embeddings in the set relating to a respective different trajectory contextual element of a user trajectory;

computing, by the processor device based on the set of embeddings, an activity representation that includes a set of POI candidate embeddings;

composing, by the processor device, a stop embedding based on the activity representation and the embeddings in the set and corresponding to a given stop in the user trajectory; and

computing, by the processor device, the trajectory-based POI recommendation using an attention-based, user-specific, multi-stop trajectory, Recurrent Neural Network (RNN) model applied to the stop embedding.

2. The computer-implemented method of claim 1, wherein said generating step comprises randomly initializing the embeddings in the set.

3. The computer-implemented method of claim 1, wherein the set of embeddings include a time embedding, a weather embedding, a static user embedding, and an item embedding.

4. The computer-implemented method of claim 1, wherein the activity representation is computed by leveraging an attention mechanism against a set of POI candidate embeddings to obtain an attention weighted average of the POI candidate embeddings as the activity representation.

5. The computer-implemented method of claim 4, wherein the attention mechanism is configured to infer a possibility of each of the POI candidate embeddings being involved in the given stop.

6. The computer-implemented method of claim 1, wherein the set of POI candidate embeddings for the given stop are predetermined from historical user data.

7. The computer-implemented method of claim, 1, wherein each of the embeddings is provided to a respective different one of a plurality of hidden layers of the RNN model for activation function processing.

8. The computer-implemented method of claim 1, wherein the stop embedding comprises an activity representation for the given stop that, in turn, comprise POI candidate embeddings.

9. The computer-implemented method of claim 1, further comprising audible and visually providing directions to the user of a POI specified by the trajectory-based POI recommendation.

10. The computer-implemented method of claim 1, further comprising automatically placing a user vehicle into an autonomous driving mode that autonomously drives the vehicle to a POI specified by the trajectory-based POI recommendation.

11. A computer program product for computing a trajectory-based Point of Interest recommendation, the computer program product comprising a non-transitory computer readable storage medium having program instructions embodied therewith, the program instructions executable by a computer to cause the computer to perform a method comprising:

generating, by a processor device, a set of embeddings, each of the embeddings in the set relating to a respective different trajectory contextual element of a user trajectory;

computing, by the processor device based on the set of embeddings, an activity representation that includes a set of POI candidate embeddings;

composing, by the processor device, a stop embedding based on the activity representation and the embeddings in the set and corresponding to a given stop in the user trajectory; and

computing, by the processor device, the trajectory-based POI recommendation using an attention-based, user-specific, multi-stop trajectory, Recurrent Neural Network (RNN) model applied to the stop embedding.

12. The computer program product of claim 11, wherein said generating step comprises randomly initializing the embeddings in the set.

13. The computer program product of claim 11, wherein the set of embeddings include a time embedding, a weather embedding, a static user embedding, and an item embedding.

14. The computer program product of claim 11, wherein the activity representation is computed by leveraging an attention mechanism against a set of POI candidate embeddings to obtain an attention weighted average of the POI candidate embeddings as the activity representation.

15. The computer program product of claim 14, wherein the attention mechanism is configured to infer a possibility of each of the POI candidate embeddings being involved in the given stop.

16. The computer program product of claim 11, wherein the set of POI candidate embeddings for the given stop are predetermined from historical user data.

17. The computer program product of claim 11, wherein each of the embeddings is provided to a respective different one of a plurality of hidden layers of the RNN model for activation function processing.

18. The computer program product of claim 11, wherein the stop embedding comprises an activity representation for the given stop that, in turn, comprise POI candidate embeddings.

19. The computer program product of claim 11, wherein the method further comprises automatically placing a user vehicle into an autonomous driving mode that autonomously drives the vehicle to a POI specified by the trajectory-based POI recommendation.

20. A computer processing system for computing a trajectory-based Point of Interest recommendation, the computer processing system comprising:

a memory for storing program code; and

a processor device for running the program code to

generate a set of embeddings, each of the embeddings in the set relating to a respective different trajectory contextual element of a user trajectory;

compute, based on the set of embeddings, an activity representation that includes a set of POI candidate embeddings;

compose a stop embedding based on the activity representation and the embeddings in the set and corresponding to a given stop in the user trajectory; and

compute the trajectory-based POI recommendation using an attention-based, user-specific, multi-stop trajectory, Recurrent Neural Network (RNN) model applied to the stop embedding.