Method and system for estimating time of arrival

Patent metadata

Field	Value
Publication number	US-10816352-B2
Application number	US-201815927073-A
Country	US
Kind code	B2
Filing date	Mar 20, 2018
Priority date	Jan 10, 2017
Publication date	Oct 27, 2020
Grant date	Oct 27, 2020

Abstract

The present disclosure relates to a method and system for determining an estimated time of arrival relating to a target trip. The method includes extracting, by a processor, sample characteristic data relating to a target trip, wherein the sample characteristic data comprises first feature data corresponding to a route relating to the target trip and second feature data corresponding to a link of the route; obtaining a prediction model for estimating time of arrival; and determining, by the processor, an estimated time of arrival (ETA) relating to the target trip based on the prediction model and the sample characteristic data.

First claim

What is claimed is:

1. A system configured to operate an online transportation service platform, comprising:
a bus;
a storage medium electronically connected to the bus, including a set of instructions for estimating time of arrival;
logic circuits in communication with the storage medium via the bus, wherein when executing the set of instructions, the logic circuits are directed to:
extract sample characteristic data relating to a target trip, wherein the sample characteristic data comprises first feature data corresponding to a route relating to the target trip and second feature data corresponding to a link of the route;
obtain a prediction model for estimating time of arrival; and
determine an estimated time of arrival (ETA) relating to the target trip based on the prediction model and the sample characteristic data. wherein, to obtain the prediction model, the logic circuits are directed to:
generate training data based on one or more historical trips; and
determine the prediction model based on the training data, wherein, to determine the prediction model, the logic circuits are directed to:
identify, from the one or more historical trips, a first plurality of training trips and a second plurality of training trips;
extract first historical characteristic data and first historical time of arrival data relating to the first plurality of training trips;
determine a first ETA determination model based on the first historical characteristic data and the first historical time of arrival;
extract second historical characteristic data and second historical time of arrival data relating to the second plurality of training trips; and
modify the first ETA determination model based on the second historical characteristic data and the second historical time of arrival data to determine a second ETA determination model.

2. The system of claim 1 , wherein the link of he route corresponds to at least a portion of the route.

3. The system of claim 1 , wherein, to determine the estimated time of arrival relating to the target trip, the logic circuits are further directed to:
determine hidden state information based on the prediction model and the second feature data; and
determine the estimated time of arrival relating to the target trip based at least in part on the hidden state information.

4. The system of claim 1 , wherein, to obtain the prediction model, the logic circuits are further directed to:
group the one or more historical trips into one or more groups based on data of links relating to the one or more historical trips, wherein the one or more historical trips are associated with a plurality of routes comprising the links;
extract historical characteristic data and historical time of arrival data for each of the one or more groups of historical trips; and
generate the training data based on the historical characteristic data and the historical time of arrival data.

5. The system of claim 1 , wherein, to determine the prediction model, the logic circuits are further directed to:
determine whether a matching condition is satisfied based on at least one of the first ETA determination model or the second ETA determination model; and
in response to determining that the matching condition is satisfied, determining the second ETA determination model as the prediction model.

6. The system of claim 5 , wherein, to determine whether the matching condition is satisfied, the logic circuits are directed to:
determine a loss function based on at least one of the first ETA determination model or the second ETA determination model; and
determine whether the loss function converges to a first value.

7. The system of claim 5 , wherein, to determine whether the matching condition is satisfied, the logic circuits are directed to:
select, from the one or more historical trips, a third plurality of historical trips;
determine an error based on third historical characteristic data and third historical time of arrival data relating to the third plurality of historical trips; and
determine whether the error is less than a second value.

8. The system of claim 1 , wherein the prediction model comprises at least one of a time series model or a regression model.

9. The system of claim 8 , wherein the time series model comprises a recurrent neural network.

10. The system of claim 8 , wherein the regression model comprises a multilayer perceptron.

11. A method configured to operate an online transportation service platform, comprising:
extracting, by a processor, sample characteristic data relating to a target trip, wherein the sample characteristic data comprises first feature data corresponding to a route relating to the target trip and second feature data corresponding to a link of the route;
obtaining, by the processor, a prediction model for estimating time of arrival; and
determining, by the processor, an estimated time of arrival (ETA) relating to the target trip based on the prediction model and the sample characteristic data, wherein obtaining the prediction model comprises:
generating training data based on one or more historical trips; and
determining the prediction model based on the training data, wherein determining the prediction model comprises:
identifying, from the one or more historical trips, a first plurality of training trips and a second plurality of training trips;
extracting first historical characteristic data and first historical time of arrival data relating to the first plurality of training trips:
determining a first ETA determination model based on the first historical characteristic data and the first historical time of arrival:
extracting second historical characteristic data and second historical time of arrival data relating to the second plurality of training trips; and
modifying the first ETA determination model based on the second historical characteristic data and the second historical time of arrival data to determine a second ETA determination model.

12. The method of claim 11 , wherein the link of the route corresponds to at least a portion of the route.

13. The method of claim 11 , wherein determining the estimated time of arrival relating to the target trip further comprises:
determining hidden state information based on the prediction model and the second feature data; and
determining the estimated time of arrival relating to the target trip based at least in part on the hidden state information.

14. The method of claim 11 , further comprising:
grouping the one or more historical trips into one or more groups based on data of links relating to the one or more historical trips, wherein the one or more historical trips are associated with a plurality of routes comprising the links;
extracting historical characteristic data and historical time of arrival data for each of the one or more groups of historical trips; and
generating the training data based on the historical characteristic data and the historical time of arrival data.

15. The method of claim 11 , further comprising:
determining whether a matching condition is satisfied based on at least one of the first ETA determination model or the second ETA determination model; and
in response to determining that the matching condition is satisfied, determining the second ETA determination model as the prediction model.

16. The method of claim 15 , wherein determining whether the matching condition is satisfied comprises:
determining a loss function based on at least one of the first ETA determination model or the second ETA determination model; and
determining whether t

Assignees

• Beijing Didi Infinity Technology & Dev Co Ltd

Inventors

• Wang Zheng
• Fu Kun

Classifications

• G06N20/00 Primary

  Machine learning · CPC title

• G01C21/3617 Primary

  using user history, behaviour, conditions or preferences, e.g. predicted or inferred from previous use or current movement · CPC title

• G06N3/048

  Activation functions · CPC title

• G06F18/214

  Generating training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title

• G01C21/3691

  Retrieval, searching and output of information related to real-time traffic, weather, or environmental conditions (arrangements for giving variable traffic instructions G08G1/09) · CPC title