Multi-source data processing method and device for urban rail transit

Abstract

The invention discloses a multi-source data processing method for urban rail transit. The multi-source data processing method comprises the following steps: S1, acquiring historical pedestrian flow data of a target station; s2, screening out a people flow sudden increase time period from the historical people flow data, and calculating a people flow sudden increase amount; s3, establishing a people flow prediction model by using the historical people flow data of which the sudden increase time period is eliminated; s4, according to the sudden increase time period, public activities in a preset range around the target station are obtained, the influence value of the public activities is calculated, and a people flow increment prediction model is established according to the influence value and the people flow increment; and S5, carrying out people flow prediction on the target site by using the people flow prediction model and the people flow burst increment prediction model. The invention further provides a multi-source data processing device. According to the invention, the accuracy of people flow prediction can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field related to data processing, and more specifically, to a multi-source data processing method and device for urban rail transit. Background Art

[0002] When predicting passenger flow at rail transit stations, existing technologies usually rely more on historical passenger flow data, and then use weather factors to make corrections. The advantage of existing technologies is that data is easy to obtain, but with the enrichment of social life, the passenger flow at rail transit stations is increasingly affected by other factors, and existing technologies have been unable to achieve accurate prediction of passenger flow. Therefore, it is necessary to integrate data from multiple sources to improve the accuracy of passenger flow prediction. Summary of the invention

[0003] An object of the present invention is to provide a multi-source data processing method and device for urban rail transit, which can improve the accuracy of passenger flow prediction.

[0004] In order to achieve these purposes and other advantages of the present invention, according to one aspect of the present invention, the present invention provides a multi-source data processing method for urban rail transit, including: S1: acquiring historical passenger flow data of a target site; S2: screening out a time period of sudden increase in passenger flow from the historical passenger flow data, and calculating a sudden increase in passenger flow; S3: establishing a passenger flow prediction model using the historical passenger flow data excluding the sudden increase time period; S4: acquiring public activities within a predetermined range around the target site based on the sudden increase time period, calculating an impact value of the public activities, and establishing a sudden increase in passenger flow prediction model based on the impact value and the sudden increase in passenger flow; S5: predicting passenger flow for the target site using the passenger flow prediction model and the sudden increase in passenger flow prediction model.

[0005] Furthermore, in the S3, it also includes: obtaining weather data and date data corresponding to the historical pedestrian flow data excluding the sudden increase time period, taking the pedestrian flow data, weather data and date data at time points T-2, T-1 and T as input, and taking the pedestrian flow data at time point T+1 as output, and using LSTM neural network training to obtain the pedestrian flow prediction model.

[0006] Furthermore, the weather data includes temperature, wind speed, haze level, rain and snow level, and the date data includes the day of the week and whether it is a holiday.

[0007] Furthermore, the method for calculating the influence value includes: obtaining the venue information and type information of the public activity, and calculating a basic influence value; obtaining posts related to the public activity corresponding to the sudden increase time period from a social platform, and calculating an increased influence value based on the posts; and calculating the influence value based on the basic influence value and the increased influence value.

[0008] Furthermore, the method for calculating the increased influence value includes: first calculating the weighted average of the number of views, the number of comments, the number of likes, and the number of reposts of the post, and then correcting it using a first correction coefficient and a second correction coefficient, wherein the first correction coefficient and the second correction coefficient are determined according to the identity of the poster and the posting address, respectively.

[0009] Furthermore, taking the added impact value and the basic impact value calculated according to each of the posts as input and the sudden increase in human flow as output, a BP neural network is trained to obtain a sudden increase in human flow prediction model.

[0010] According to another aspect of the present invention, a multi-source data processing device for urban rail transit is also provided, including: an acquisition module for acquiring historical passenger flow data of a target site; a calculation module for screening out a time period of sudden passenger flow increase from the historical passenger flow data, and calculating the sudden passenger flow increase; a first establishment module for establishing a passenger flow prediction model using the historical passenger flow data excluding the sudden passenger flow time period; a second establishment module for acquiring public activities within a predetermined range around the target site according to the sudden passenger flow time period, calculating the impact value of the public activities, and establishing a sudden passenger flow increase prediction model according to the impact value and the sudden passenger flow increase; a prediction module for predicting passenger flow for the target site using the passenger flow prediction model and the sudden passenger flow increase prediction model.

[0011] Furthermore, the first establishment module is also used to obtain the weather data and date data corresponding to the historical passenger flow data excluding the sudden increase time period, and take the passenger flow data, weather data and date data at time points T-2, T-1 and T as input, and the passenger flow data at time point T+1 as output, and use LSTM neural network training to obtain the passenger flow prediction model; the weather data includes temperature, wind speed, haze level, rain and snow level, and the date data includes the day of the week and whether it is a holiday.

[0012] Furthermore, the method for calculating the impact value includes: obtaining the venue information and type information of the public activity, and calculating a basic impact value; obtaining posts related to the public activity corresponding to the sudden increase time period from a social platform, and calculating an increased impact value based on the posts; the method for calculating the increased impact value includes: first calculating the weighted average of the number of views, the number of comments, the number of likes, and the number of reposts of the post, and then correcting it using a first correction coefficient and a second correction coefficient, wherein the first correction coefficient and the second correction coefficient are respectively determined according to the identity of the poster and the posting address.

[0013] Furthermore, taking the added impact value and the basic impact value calculated according to each of the posts as input and the sudden increase in human flow as output, a BP neural network is trained to obtain a sudden increase in human flow prediction model.

[0014] The present invention has at least the following beneficial effects: The present invention obtains historical pedestrian flow data of a target site, filters out a time period of sudden increase in pedestrian flow from the historical pedestrian flow data, calculates the sudden increase in pedestrian flow, establishes a pedestrian flow prediction model by using the historical pedestrian flow data after excluding the sudden increase time period, obtains public activities within a predetermined range around the target site according to the sudden increase time period, calculates the impact value of the public activities, establishes a sudden increase in pedestrian flow prediction model according to the impact value and the sudden increase in pedestrian flow, and predicts the pedestrian flow of the target site by using the pedestrian flow prediction model and the sudden increase in pedestrian flow prediction model; the present invention not only takes into account the historical pedestrian flow data, but also takes into account the impact of public activities on the sudden increase in pedestrian flow, thereby improving the accuracy of pedestrian flow prediction.

[0015] Other advantages, objectives and features of the present invention will be embodied in part through the following description, and in part will be understood by those skilled in the art through study and practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a flow chart of an embodiment of the present application. DETAILED DESCRIPTION

[0017] The present invention is further described in detail below in conjunction with the accompanying drawings so that those skilled in the art can implement the invention with reference to the description.

[0018] It should be understood that the terms such as "having", "including" and "comprising" used in the embodiments of the present application do not exclude the existence or addition of one or more other elements or their combinations. All directional indications (such as up, down, left, right, front, back...) in the embodiments of the present application are only used to explain the relative position relationship, movement, etc. between the components in a certain specific posture. If the specific posture changes, the directional indication will also change accordingly. When an element is referred to as "fixed on" or "set on" another element, it can be directly on the other element or there may be a centering element at the same time. When an element is referred to as "connecting" another element, it can be directly connected to another element or it can be indirectly connected to another element through a centering element. The description of "first", "second", etc. in the embodiments of the present application is only for descriptive purposes, and cannot be understood as indicating or implying its relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" can explicitly or implicitly include at least one of the features.

[0019] It should be noted that the technical solutions between the various embodiments of the present application can be combined with each other, but it must be based on the fact that ordinary technicians in the field can implement it. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by this application.

[0020] like Figure 1 As shown, an embodiment of the present application provides a multi-source data processing method for urban rail transit, including: S1: Obtain historical traffic data of the target site; For example, the flow of people in each historical time period can be determined by the number of card swipes at the target station. The length of the time period can be determined according to specific forecasting needs, which can be 1 day, 3 days, 7 days, 10 days, etc.; Exemplarily, the target site may be a rail transit site or two to three rail transit sites that are close in distance; S2: Filtering out a time period of sudden increase in passenger flow from the historical passenger flow data, and calculating a sudden increase in passenger flow; Exemplarily, the average passenger flow is determined based on the historical passenger flow, the sudden increase time period is screened out based on the average passenger flow, and the sudden increase of passenger flow is calculated based on the difference between the passenger flow in the sudden increase time period and the average passenger flow; Exemplarily, a flow rate change image is established with the flow rate and time point coordinate axes, and the sudden increase time period is determined according to the slope change in the flow rate change image; S3: Establishing a crowd flow prediction model by using the historical crowd flow data excluding the sudden increase time period; In this step, the sudden increase period is first eliminated, and the remaining data does not have abnormal fluctuations, so the established passenger flow prediction model is more accurate; Exemplarily, a pedestrian flow prediction model is established using regression methods, machine learning, neural networks, etc.; S4: acquiring public activities within a predetermined range around the target site according to the sudden increase time period, calculating the impact value of the public activities, and establishing a sudden increase prediction model for human flow according to the impact value and the sudden increase of human flow; In this step, it can be understood that the time of holding the public event can be the same as the sudden increase time period or there can be a certain overlap, so as to fully consider the impact of the public event on the sudden increase in passenger flow; Exemplarily, a public event may be an exhibition, a concert, a competition, a food festival, a business opening, etc.; For example, the impact value is assigned according to the level and popularity of the public activity, and then a sudden increase prediction model of the flow of people is established by using regression methods, machine learning, neural networks, etc.; Exemplarily, the predetermined range is an area that is easily accessible by foot or bicycle, such as within a range of 5 kilometers; S5: predicting the flow of people at the target site using the flow prediction model and the sudden increase in flow prediction model; For example, when predicting the flow of people, the general flow of people is predicted according to the flow of people prediction model, the public activities in the corresponding time period are obtained, and the sudden increase flow of people is predicted using the sudden increase flow of people prediction model. The general flow of people and the sudden increase flow of people are added together to obtain the predicted flow of people; It can be seen that compared with the existing technology, this embodiment not only takes into account the historical passenger flow data, but also takes into account the impact of public activities on the sudden increase in passenger flow. It uses multi-source data fusion to improve the accuracy of passenger flow prediction and overcome the defects of the existing technology.

[0021] In another embodiment, in S3, it further includes: obtaining weather data and date data corresponding to the historical pedestrian flow data excluding the sudden increase time period, taking the pedestrian flow data, weather data and date data at time points T-2, T-1 and T as input, taking the pedestrian flow data at time point T+1 as output, and using LSTM neural network training to obtain the pedestrian flow prediction model; This embodiment provides a method for establishing a passenger flow prediction model using an LSTM neural network. LSTM is more suitable for predicting time series. In this embodiment, T represents a time period, T-2, T-1 and T+1 represent the first two time periods, the previous time period and the next time period respectively. The weather data mainly includes weather data such as temperature, rain and snow that affect travel willingness. The date data includes date data that may affect travel. The historical passenger flow data, weather data and date data excluding the sudden increase time period are used to establish a training set. The weather data and date data are preprocessed and formed into vectors with the passenger flow data. The LSTM network is trained to obtain a passenger flow prediction model. Exemplarily, the weather data includes temperature, wind speed, haze level, rain and snow level, and the date data includes the day of the week and whether it is a holiday. The above weather data and date data are digitized; Exemplarily, the LSTM includes an input layer, an LSTM layer and an output layer, the number of training samples is no less than 100, the mean absolute error loss function is adopted, the input layer is 7-dimensional, the output layer is 1-dimensional, and the number of hidden layer neural units is 20.

[0022] In another embodiment, the method for calculating the influence value includes: obtaining the venue information and type information of the public activity, and calculating the basic influence value; obtaining posts related to the public activity corresponding to the sudden increase time period from a social platform, and calculating the increased influence value according to the posts; This embodiment provides a method for calculating the influence value, according to the venue information and type information of the public activity, including sub-items such as the distance between the venue and the target subway station, the area of ​​the venue, and the level of the public activity, the basic influence value is calculated using the score of each sub-item, wherein the distance between the venue and the target subway station and the area of ​​the venue can be determined by expert scoring, and the average value is taken; the level of the public activity can be determined according to the information of the organizer of the activity, such as the government, school, social group, merchant, etc., and different scores are assigned respectively; it can also be determined according to the information of the participating characters, such as assigning different scores according to the number of fans of the participating characters on the social platform, and the scores of each sub-item are added up to obtain the basic influence value; Exemplarily, the method for calculating the increased influence value includes: first calculating a weighted average of the number of views, the number of comments, the number of likes, and the number of reposts of the post, and then correcting the weighted average using a first correction coefficient and a second correction coefficient, wherein the first correction coefficient and the second correction coefficient are determined according to the identity of the poster and the posting address, respectively; Exemplarily, the posts related to the public activity corresponding to the sudden increase time period refer to posts in a period of time before the sudden increase time period, such as within 1-10 days before the sudden increase time period; For example, the post includes not only text but also videos, images, etc. The social platform may be Weibo, Xiaohongshu, Douyin, etc.; Exemplarily, the added influence value s=α*β*(n1*a+n2*b+n3*c+n4*d), a, b, c and d represent the number of views, the number of comments, the number of likes and the number of reposts respectively, n1, n2, n3 and n4 are weights respectively, α and β are the first correction coefficient and the second correction coefficient respectively; n1, n2, n3 and n4 can adopt empirical values, generally the number of comments has the highest weight, the number of likes and the number of reposts are second, and the number of views is the lowest, and the hierarchical analysis method can also be used to determine; α and β are used to correct the added influence value, and the values ​​are between 0-1, which are used to reduce the interference of invalid posts so that the influence value can truly represent the popularity of public activities. The identity of the poster can be the social platform account level of the poster, and the posting address can be the IP address of the poster. Generally speaking, the higher the level of the poster's social platform account, the larger the α value, and the closer the poster's IP address is to the target subway station, the larger the β value. The values ​​of α and β can be determined after analyzing the actually obtained data, and this embodiment does not make specific limitations.

[0023] In another embodiment, the added impact value and the basic impact value calculated according to each of the posts are used as input, and the sudden increase in human flow is used as output to train a BP neural network to obtain a sudden increase in human flow prediction model.

[0024] This embodiment provides a specific model for predicting sudden increase in traffic flow, which is obtained by inputting the increase impact value and basic impact value of each post calculated in the previous embodiment into a BP neural network, and taking the sudden increase in traffic flow as output, and training; Exemplarily, each post is sorted according to the increased influence value, and the top 20-30% of the increased influence values ​​are selected to further reduce the interference of low-quality posts, and together with the basic influence value, they form an input vector as the input of the BP neural network; Exemplarily, the BP neural network includes an input layer, a hidden layer and an output layer, the input layer has 50 nodes, the output layer has 1 node, the hidden layer has 80 nodes, the learning rate is set to 0.0003, and the performance function is the mean square error MSE; After testing, the passenger flow prediction model trained with the LSTM network and the passenger flow sudden increase prediction model trained with the BP neural network have a 27% higher passenger flow prediction accuracy than the prediction model trained with the polynomial regression model (taking historical passenger flow, weather, and day of the week into consideration).

[0025] Based on the same inventive concept, an embodiment of the present application further provides a multi-source data processing device for urban rail transit, including: an acquisition module, used to acquire historical passenger flow data of a target site; a calculation module, used to screen out a passenger flow surge time period from the historical passenger flow data, and calculate a passenger flow surge increment; a first establishment module, used to establish a passenger flow prediction model using the historical passenger flow data excluding the surge time period; a second establishment module, used to acquire public activities within a predetermined range around the target site according to the surge time period, calculate the impact value of the public activities, and establish a passenger flow surge increment prediction model according to the impact value and the passenger flow surge increment; a prediction module, used to predict passenger flow for the target site using the passenger flow prediction model and the passenger flow surge increment prediction model; This embodiment uses a computer program to construct an acquisition module, a calculation module, a first establishment module, a second establishment module and a prediction module. By establishing a crowd flow prediction model and a crowd flow sudden increase prediction model, it not only takes into account the historical crowd flow data, but also takes into account the impact of public activities on the sudden increase of crowd flow. That is, it uses multi-source data fusion to improve the accuracy of crowd flow prediction and overcome the defects of the existing technology.

[0026] In another embodiment, the first establishing module is also used to obtain the weather data and date data corresponding to the historical pedestrian flow data excluding the sudden increase time period, and take the pedestrian flow data, weather data and date data at time points T-2, T-1 and T as input, and take the pedestrian flow data at time point T+1 as output, and use LSTM neural network training to obtain the pedestrian flow prediction model; the weather data includes temperature, wind speed, haze level, rain and snow level, and the date data includes the day of the week and whether it is a holiday; this embodiment provides a method for establishing a pedestrian flow prediction model using LSTM neural network, LSTM is more suitable for time series prediction, T-2, T-1 and T+1 respectively represent the first two time periods, the previous time period and the next time period, and use the historical pedestrian flow data, weather data and date data excluding the sudden increase time period to establish a training set, the weather data and date data are preprocessed and the pedestrian flow data are formed into vectors, the LSTM network is trained, and the pedestrian flow prediction model is obtained.

[0027] In another embodiment, the method for calculating the influence value includes: obtaining the venue information and type information of the public activity, and calculating a basic influence value; obtaining posts related to the public activity corresponding to the sudden increase time period from a social platform, and calculating an increased influence value based on the posts; the method for calculating the increased influence value includes: first calculating the weighted average of the number of views, the number of comments, the number of likes, and the number of reposts of the post, and then correcting it using a first correction coefficient and a second correction coefficient, wherein the first correction coefficient and the second correction coefficient are determined according to the identity of the poster and the posting address, respectively; this embodiment provides a second establishment module for calculating the influence value, according to the venue information and type information of the public activity, including sub-items such as the distance between the venue and the target subway station, the venue area, and the level of the public activity, the basic influence value is calculated using the score of each sub-item, the score of each sub-item is added to obtain the basic influence value, and posts are obtained from the social platform, and the increased influence value is calculated based on the interaction data.

[0028] In another embodiment, the increased influence value and the basic influence value calculated according to each of the posts are used as input, and the sudden increase in human flow is used as output to train a BP neural network to obtain a sudden increase in human flow prediction model; that is, the increased influence value and the basic influence value of each post calculated in the previous embodiment are input into the BP neural network, and the sudden increase in human flow is used as output to obtain the training model.

[0029] Although the embodiments of the present invention have been disclosed as above, they are not limited to the applications listed in the specification and the implementation modes, and they can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and the illustrations shown and described herein.

Claims

1. A multi-source data processing method for urban rail transit, characterized in that: include: S1: Obtain historical traffic data of the target site; S2: Filtering out a time period of sudden increase in passenger flow from the historical passenger flow data, and calculating a sudden increase in passenger flow; S3: Establishing a crowd flow prediction model by using the historical crowd flow data excluding the sudden increase time period; S4: acquiring public activities within a predetermined range around the target site according to the sudden increase time period, calculating the impact value of the public activities, and establishing a sudden increase prediction model for human flow according to the impact value and the sudden increase of human flow; S5: Utilize the passenger flow prediction model and the passenger flow sudden increase prediction model to predict passenger flow at the target site.

2. The multi-source data processing method for urban rail transit according to claim 1, characterized in that: In the S3, it also includes: The weather data and date data corresponding to the historical pedestrian flow data excluding the sudden increase time period are obtained, the pedestrian flow data, weather data and date data at time points T-2, T-1 and T are used as input, and the pedestrian flow data at time point T+1 is used as output, and the pedestrian flow prediction model is obtained by training using an LSTM neural network.

3. The multi-source data processing method for urban rail transit according to claim 2, characterized in that: The weather data includes temperature, wind speed, haze level, rain and snow level, and the date data includes the day of the week and whether it is a holiday.

4. The multi-source data processing method for urban rail transit according to claim 2, characterized in that: The calculation method of the impact value includes: Obtaining the venue information and type information of the public activity, and calculating the basic impact value; Posts related to the public activity corresponding to the sudden increase time period are obtained from a social platform, and an increased impact value is calculated according to the posts.

5. The multi-source data processing method for urban rail transit according to claim 4, characterized in that: The method for calculating the increased influence value includes: first calculating the weighted average of the number of views, the number of comments, the number of likes, and the number of reposts of the post, and then correcting it using a first correction coefficient and a second correction coefficient, wherein the first correction coefficient and the second correction coefficient are determined according to the identity of the poster and the posting address, respectively.

6. The multi-source data processing method for urban rail transit according to claim 5, characterized in that: The increased impact value and the basic impact value calculated according to each of the posts are used as input, and the sudden increase in human flow is used as output to train a BP neural network to obtain a sudden increase in human flow prediction model.

7. A multi-source data processing device for urban rail transit, characterized in that: include: An acquisition module is used to obtain historical traffic data of a target site; A calculation module, which selects a time period of sudden increase in passenger flow from the historical passenger flow data and calculates a sudden increase in passenger flow; A first establishing module is used to establish a crowd flow prediction model by using the historical crowd flow data excluding the sudden increase time period; A second establishment module is used to obtain public activities within a predetermined range around the target site according to the sudden increase time period, calculate the impact value of the public activities, and establish a sudden increase prediction model for human flow according to the impact value and the sudden increase of human flow; The prediction module uses the passenger flow prediction model and the passenger flow sudden increment prediction model to predict the passenger flow of the target site.

8. The multi-source data processing device for urban rail transit according to claim 7, characterized in that: The first establishment module is also used to obtain the weather data and date data corresponding to the historical passenger flow data excluding the sudden increase time period, and take the passenger flow data, weather data and date data at time points T-2, T-1 and T as input, and the passenger flow data at time point T+1 as output, and use LSTM neural network training to obtain the passenger flow prediction model; the weather data includes temperature, wind speed, haze level, rain and snow level, and the date data includes the day of the week and whether it is a holiday.

9. The multi-source data processing device for urban rail transit according to claim 8, characterized in that: The method for calculating the impact value includes: obtaining the venue information and type information of the public activity, and calculating a basic impact value; obtaining posts related to the public activity corresponding to the sudden increase time period from a social platform, and calculating an increased impact value based on the posts; the method for calculating the increased impact value includes: first calculating a weighted average of the number of views, the number of comments, the number of likes, and the number of reposts of the post, and then correcting it using a first correction coefficient and a second correction coefficient, wherein the first correction coefficient and the second correction coefficient are respectively determined according to the identity of the poster and the posting address.

10. The multi-source data processing device for urban rail transit according to claim 9, characterized in that: The increased impact value and the basic impact value calculated according to each of the posts are used as input, and the sudden increase in human flow is used as output to train a BP neural network to obtain a sudden increase in human flow prediction model.

Images