CN-121981368-A - Passenger flow prediction method, device, equipment and medium

CN121981368ACN 121981368 ACN121981368 ACN 121981368ACN-121981368-A

Abstract

The present application relates to the field of big data analysis technologies, and in particular, to a passenger flow prediction method, device, equipment, and medium. According to the method, the travel utility model and the passenger flow prediction model are respectively trained based on the data of the competitive operation line to form a mixed model for prediction, the passenger flow is predicted by using the mixed model, the advantages of traffic selection theory and machine learning are fused, decoupling analysis of the multi-factor dynamic relationship of the passenger flow is realized, and the accuracy of passenger flow prediction is improved.

Inventors

TIAN XIANCAI
WU JINYONG
YU XIAOTIAN
LI AIJUN

Assignees

深圳云天励飞技术股份有限公司
青岛云天励飞科技有限公司

Dates

Publication Date: 20260505
Application Date: 20251212

Claims (10)

1. A passenger flow prediction method, comprising: Obtaining planning operation data of a target operation line to be analyzed and actual operation data of an existing operation line, comparing the planning operation data with the actual operation data to obtain a comparison result, and determining a competition operation line from the existing operation line according to the comparison result; Training a preset travel utility model by using actual operation data of the competition operation line to obtain a trained travel utility model, inputting the actual operation data into the trained travel utility model, and outputting a first predicted passenger flow; Taking the actual running data and the first predicted passenger flow as a group of training input data, taking the actual passenger flow of the competition running line as a label corresponding to the training input data, and training a preset passenger flow prediction model to obtain a trained passenger flow prediction model; And inputting the planning operation data into the trained travel utility model, outputting a second predicted passenger flow, inputting the planning operation data and the second predicted passenger flow into the trained passenger flow prediction model, and outputting a target predicted passenger flow.
2. The passenger flow prediction method according to claim 1, further comprising, after the inputting the planned operation data and the second predicted passenger flow into a trained passenger flow prediction model and outputting a target predicted passenger flow: Determining the passenger flow loss corresponding to the competitive operation line according to the target predicted passenger flow and the actual passenger flow, and updating the actual operation data according to the passenger flow loss of the competitive operation line to obtain updated operation data; Inputting the updated operation data and the first predicted passenger flow into the trained passenger flow prediction model to obtain a third predicted passenger flow, and updating the target predicted passenger flow according to the third predicted passenger flow to obtain an updated target predicted passenger flow; Taking the updated target predicted passenger flow as the target predicted passenger flow, returning to execute the passenger flow loss corresponding to the competitive operation line according to the target predicted passenger flow and the actual passenger flow; and when the iteration conditions are satisfied in the loop execution, determining the updated target predicted passenger flow corresponding to the iteration conditions as the final predicted passenger flow corresponding to the target operation line.
3. The passenger flow prediction method according to claim 2, wherein determining, when the loop execution satisfies an iteration condition, the updated target predicted passenger flow corresponding to the iteration condition as the final predicted passenger flow corresponding to the target running line includes: Collecting third predicted passenger flows and updated target predicted passenger flows in two continuous iterations; calculating a first change rate of a third predicted passenger flow in two successive iterations for the competitive operation line, and calculating a second change rate of an updated target predicted passenger flow in two successive iterations for the target operation line; And when the first change rate and the second change rate are smaller than a preset change rate threshold, determining that the loop execution meets an iteration condition, and randomly selecting an updated target predicted passenger flow from two continuous iterations as the final predicted passenger flow corresponding to the target running line.
4. The passenger flow prediction method according to claim 1, wherein training a preset travel utility model by using actual operation data of the competitive operation line to obtain a trained travel utility model comprises: analyzing the actual operation data, determining at least one service attribute, constructing a utility function based on all the service attributes, and obtaining a preset travel utility model by combining multiple logistic regression based on the utility function, wherein each service attribute in the utility function corresponds to a weight parameter; analyzing the actual operation data, determining an attribute value of each service attribute of each initial end point pair, and determining a passenger flow distribution probability corresponding to each initial end point pair according to the actual passenger flows of all competing operation lines; And training the weight parameters in the preset travel utility model according to the attribute value and the passenger flow distribution probability of each service attribute of any initial and final point pair to obtain a trained travel utility model.
5. The passenger flow prediction method according to claim 4, wherein the inputting the actual operation data into the trained travel utility model and outputting a first predicted passenger flow comprises: for any initial end point pair, acquiring the total amount of passenger flow of all competing running lines on the initial end point pair, and extracting attribute values of each service attribute on the initial end point pair from the actual running data; Inputting the attribute value of each service attribute into the trained travel utility model, and outputting a corresponding first allocation probability; And calculating to obtain a first predicted passenger flow according to the passenger flow total amount and the first allocation probability.
6. The passenger flow prediction method according to claim 1, wherein the training the preset passenger flow prediction model to obtain a trained passenger flow prediction model comprises: carrying out passenger flow residual prediction on actual operation data in input training input data by using a preset passenger flow prediction model to obtain predicted residual passenger flow; Summing the predicted residual passenger flow and the first predicted passenger flow in the training input data to obtain a summation result, and determining a predicted loss according to the summation result and the actual passenger flow in the training input data; And training model parameters of the preset passenger flow prediction model according to the prediction loss to obtain a trained passenger flow prediction model.
7. The passenger flow prediction method according to any one of claims 1 to 6, wherein the comparing the planned operation data with the actual operation data to obtain a comparison result, and determining a competitive operation line from existing operation lines according to the comparison result, includes: analyzing the planning operation data, and determining the geographic coordinates of each starting and ending point pair in the target operation line, the physical path of the target operation line and the operation time table of the target operation line; analyzing the actual operation data, and determining the geographic coordinates of each starting and ending point pair in the existing operation line, the physical path of the existing operation line and the operation schedule of the existing operation line; calculating the similarity between the geographic coordinates of any initial end point pair of the target operation line and the geographic coordinates of any initial end point pair of the existing operation line to obtain the similarity of the initial end point pair; Calculating the overlapping degree of the physical path of the target operation line and the physical path of the existing operation line to obtain the path overlapping degree; Calculating the overlapping degree of the operation time table of the target operation line and the operation time table of the existing operation line to obtain service time overlapping degree; and carrying out weighted fusion on the similarity, the path overlapping degree and the service time overlapping degree by the starting and ending points to obtain an overlapping degree competition index, and determining that the existing operation line with the overlapping degree competition index being greater than a preset index threshold is a competition operation line.
8. A passenger flow prediction device, characterized by comprising: The competition route determining module is used for acquiring planning operation data of a target operation line to be analyzed and actual operation data of an existing operation line, comparing the planning operation data with the actual operation data to obtain a comparison result, and determining the competition operation line from the existing operation line according to the comparison result; The first training module is used for training a preset travel utility model by using actual operation data of the competition operation line to obtain a trained travel utility model, inputting the actual operation data into the trained travel utility model, and outputting a first predicted passenger flow; The second training module is used for training a preset passenger flow prediction model by taking the actual running data and the first predicted passenger flow as a group of training input data and taking the actual passenger flow of the competition running line as a label of the corresponding training input data to obtain a trained passenger flow prediction model; The passenger flow prediction module is used for inputting the planning operation data into the trained travel utility model, outputting a second predicted passenger flow, inputting the planning operation data and the second predicted passenger flow into the trained passenger flow prediction model, and outputting a target predicted passenger flow.
9. A computer device, characterized in that it comprises a processor, a memory and a computer program stored in the memory and executable on the processor, which processor implements the passenger flow prediction method according to any one of claims 1 to 7 when executing the computer program.
10. A computer readable storage medium storing a computer program, characterized in that the computer program, when executed by a processor, implements the passenger flow prediction method according to any one of claims 1 to 7.

Description

Passenger flow prediction method, device, equipment and medium Technical Field The present application relates to the field of big data analysis technologies, and in particular, to a passenger flow prediction method, device, equipment, and medium. Background In the field of smart city and public transportation planning, the method for predicting the passenger flow of a new public transportation route scientifically and accurately is a key link of route planning, capacity allocation and financial subsidy assessment. Conventional passenger flow prediction methods generally rely on macroscopic data such as regional population density, land utilization properties, resident trip surveys and the like, and are estimated by adopting an attraction model or a conventional four-stage method. These methods suffer from the following significant drawbacks: 1) The prediction accuracy is insufficient, that is, the traditional method is mostly based on static and macroscopic data, so that complex decision behaviors of passengers facing a plurality of optional routes are difficult to capture, and the direct competition relationship among the routes is ignored, namely, how the new routes attract passenger flows from the existing routes by virtue of service advantages (such as lower fare and shorter travel time). 2) The existing method can only give a rough estimation of total passenger flow in all days, and cannot provide passenger flow distribution conditions of different periods such as peaks, flat peaks and the like, so that the prediction result has limited guiding significance for fine operation scheduling (such as dynamic adjustment of departure frequency). 3) The sensitivity to service attributes is low, and the direct influence of specific service attribute changes such as fare, departure frequency, riding comfort and the like on passenger flow is difficult to quantify by the traditional model, so that the reliable data basis is lacked when line optimization and scheme comparison are carried out. Therefore, how to perform decoupling analysis on the multi-factor dynamic relationship of the passenger flow to optimize the passenger flow prediction flow, so that improving the accuracy of passenger flow prediction becomes a problem to be solved urgently. Disclosure of Invention In view of the above, the embodiments of the present application provide a method, apparatus, device, and medium for predicting a passenger flow, so as to solve the problem of how to perform decoupling analysis on a multi-factor dynamic relationship of the passenger flow to optimize a passenger flow prediction flow, thereby improving accuracy of passenger flow prediction. In a first aspect, an embodiment of the present application provides a passenger flow prediction method, including: Obtaining planning operation data of a target operation line to be analyzed and actual operation data of an existing operation line, comparing the planning operation data with the actual operation data to obtain a comparison result, and determining a competition operation line from the existing operation line according to the comparison result; Training a preset travel utility model by using actual operation data of the competition operation line to obtain a trained travel utility model, inputting the actual operation data into the trained travel utility model, and outputting a first predicted passenger flow; Taking the actual running data and the first predicted passenger flow as a group of training input data, taking the actual passenger flow of the competition running line as a label corresponding to the training input data, and training a preset passenger flow prediction model to obtain a trained passenger flow prediction model; And inputting the planning operation data into the trained travel utility model, outputting a second predicted passenger flow, inputting the planning operation data and the second predicted passenger flow into the trained passenger flow prediction model, and outputting a target predicted passenger flow. In a second aspect, an embodiment of the present application provides a passenger flow prediction apparatus, including: The competition route determining module is used for acquiring planning operation data of a target operation line to be analyzed and actual operation data of an existing operation line, comparing the planning operation data with the actual operation data to obtain a comparison result, and determining the competition operation line from the existing operation line according to the comparison result; The first training module is used for training a preset travel utility model by using actual operation data of the competition operation line to obtain a trained travel utility model, inputting the actual operation data into the trained travel utility model, and outputting a first predicted passenger flow; The second training module is used for training a preset passenger flow prediction model by taking the actual running data a