Route prediction based on adaptive hybrid model

1. A computer-implemented method, comprising:

obtaining a first route traversed by a target object;

training an object-specific prediction model using history route data of the target object;

performing at least one prediction for a second route to be traversed by the target object based on a hybrid model jointly applying the first route, the at least one prediction being performed with a hybrid consideration using the hybrid model of the object-specific prediction model, an object group-specific prediction model, and an object-independent prediction model; and

determining, according to a decision rule, a prediction result of the second route based on the at least one prediction,

wherein the hybrid model completes the prediction when history data is missing or insufficient from a beginning phase of a route prediction platform from one of the object-specific prediction model, the object group-specific prediction model, and the object-independent prediction model by running computations of the prediction through the object-independent prediction model of the hybrid model instead of the object-specific prediction model and the object group-specific prediction model, and

wherein the performing performs the prediction based on the hybrid model without requiring history data.

2. The method of claim 1 , wherein the determining the prediction result comprises:

obtaining respective intermediate prediction results of the at least one prediction using testing data that is provided to the hybrid model; and

determining the prediction result from the intermediate prediction results.

3. The method of claim 2 , wherein the determining the prediction result from the intermediate prediction results comprises one of:

selecting one of the intermediate prediction results as the prediction result; and

combining two or more of the intermediate prediction results into the prediction result.

4. The method of claim 1 , wherein the object-specific prediction model includes a pattern matching model, the object group-specific prediction model includes a deep learning model, and the object-independent prediction model includes a road network-based model.

5. The method of claim 1 , wherein the decision rule is based on at least one of:

a credibility parameter of a respective one of the prediction models representing credibility of the respective prediction model; and

a route profile of the target object.

6. The method of claim 5 , wherein the credibility parameter includes at least one of:

a history count of a destination or a matching degree of an intermediate prediction result of the object-specific prediction model; and

a prediction probability of an intermediate prediction result of the object group-specific prediction model; or

a route efficiency of an intermediate prediction result of the object-independent prediction model.

7. The method of claim 1 , further comprising at least one of:

training the object group-specific prediction model using history route data of an object group; and

training the decision rule using history route data of at least one of the target object and the object group.

8. The method of claim 1 , wherein the determining the prediction result comprises:

obtaining respective intermediate prediction results of each part of the hybrid model including the object-specific prediction model, the object group-specific prediction model, and the object-independent prediction model using testing data that is provided to the hybrid model;

determining the prediction result from the intermediate prediction results; and

training the decision rule according to intermediate prediction results.

9. A system comprising:

one or more processors;

a memory coupled to at least one of the processors; and

a set of computer program instructions stored in the memory and executed by at least one of the processors in order to perform actions of:

obtaining a first route traversed by a target object;

training an object-specific prediction model using history route data of the target object;

performing at least one prediction for a second route to be traversed by the target object based on a hybrid model jointly applying the first route, the at least one prediction being performed with a hybrid consideration using the hybrid model of gr object-specific prediction model, an object group-specific prediction model, and an object-independent prediction model; and

determining, according to a decision rule, a prediction result of the second route based on the at least one prediction,

wherein the hybrid model completes the prediction when history data is missing or insufficient from a beginning phase of a route prediction platform from one of the object-specific prediction model, the object group-specific prediction model, and the object-independent prediction model by running computations of the prediction through the object-independent prediction model of the hybrid model instead of the object-specific prediction model and the object group-specific prediction model, and

wherein the performing performs the prediction based on the hybrid model without requiring history data.

10. The system of claim 9 , wherein the determining the prediction result comprises:

obtaining respective intermediate prediction results of the at least one prediction; and

determining the prediction result from the intermediate prediction results.

11. The system of claim 10 , wherein the determining the prediction result from the intermediate prediction results comprises one of:

selecting one of the intermediate prediction results as the prediction result; and

combining two or more of the intermediate prediction results into the prediction result.

12. The system of claim 9 , wherein the object-specific prediction model includes a pattern matching model, the object group-specific prediction model includes a deep learning model, and the object-independent prediction model includes a road network-based model.

13. The system of claim 9 , wherein the decision rule is based on at least one of:

a credibility parameter of a respective one of the prediction models representing credibility of the respective prediction model; and

a route profile of the target object.

14. The system of claim 13 , wherein the credibility parameter includes at least one of:

a history count of a destination or a matching degree of an intermediate prediction result of the object-specific prediction model;

a prediction probability of an intermediate prediction result of the object group-specific prediction model; and

a route efficiency of an intermediate prediction result of the object-independent prediction model.

15. The system of claim 9 , wherein the actions further include at least one of:

training the object group-specific prediction model using history route data of an object group; and

training the decision rule using history route data of at least one of the target object or the object group.

16. A computer program product being tangibly stored on a non-transitory machine-readable medium and comprising machine-executable instructions, the instructions, when executed on a device, causing the device to perform actions of:

obtaining a first route traversed by a target object;

training an object-specific prediction model using history route data of the target object;

performing at least one prediction for a second route to be traversed by the target object based on a hybrid model jointly applying the first route, the at least one prediction being performed with a hybrid consideration using the hybrid model of the object-specific prediction model, an object group-specific prediction model, and an object-independent prediction model; and

determining, according to a decision rule, a prediction result of the second route based on the at least one prediction,

wherein the hybrid model completes the prediction when history data is missing or insufficient from a beginning phase of a route prediction platform from one of the object-specific prediction model, the object group-specific prediction model, and the object-independent prediction model by running computations of the prediction through the object-independent prediction model of the hybrid model instead of the object-specific prediction model and the object group-specific prediction model, and

wherein the performing performs the prediction based on the hybrid model without requiring history data.

17. The computer program product of claim 16 , wherein the determining the prediction result comprises:

obtaining respective intermediate prediction results of the at least one prediction; and

determining the prediction result from the intermediate prediction results.

18. The computer program product of claim 17 , wherein the determining the prediction result from the intermediate prediction results comprises one of:

selecting one of the intermediate prediction results as the prediction result; and

combining two or more of the intermediate prediction results into the prediction result.

19. The computer program product of claim 16 , wherein the decision rule is based on each of:

a credibility parameter of a respective one of the prediction models representing credibility of the respective prediction model; and

a route profile of the target object.

20. The computer program product of claim 19 , wherein the credibility parameter includes each of:

a history count of a destination or a matching degree of an intermediate prediction result of the object-specific prediction model;

a prediction probability of an intermediate prediction result of the object group-specific prediction model; and

a route efficiency of an intermediate prediction result of the object-independent prediction model.