CN107180278A

CN107180278A - A kind of real-time passenger flow forecasting of track traffic

Info

Publication number: CN107180278A
Application number: CN201710387638.9A
Authority: CN
Inventors: 杨梦宁; 许任婕; 李小斌; 赵小超; 徐玲; 葛永新; 洪明坚; 黄晟; 王洪星; 陈飞宇
Original assignee: Chongqing University
Current assignee: Chongqing University
Priority date: 2017-05-27
Filing date: 2017-05-27
Publication date: 2017-09-19

Abstract

The present invention relates to a kind of real-time passenger flow forecasting of track traffic, low, the computationally intensive technical problem of predictablity rate present in prior art is mainly solved, the present invention includes by using method：N historical data is gathered as original sample from AFC system, and original sample progress is pre-processed and obtains pre-processing sample；Sample is pre-processed according in, the Passenger flow forecast model in short-term based on support vector regression is set up according to kernel function and fitting regression function, the kernel function is RBF；Anticipation function is fitted using time series vector X and with time series to the corresponding passenger flow vector Y of X amounts as input, the input is fitted anticipation function input step Passenger flow forecast model, passenger flow forecast vector Y in short-term_n+1；Technical scheme, the problem is preferably resolved, available in the real-time passenger flow estimation of track traffic.

Description

Real-time passenger flow prediction method for rail transit

Technical Field

The invention relates to the field of rail transit passenger flow prediction, in particular to a rail transit real-time passenger flow prediction method.

Background

With the rapid development of the urbanization process, the conflict between the travel demand of urban population and the urban traffic capacity is more prominent. Urban rail transit is distinguished from various traffic modes by the advantages of high speed, high capacity, environmental protection and the like, and becomes a main vehicle for solving traffic jam. The urban rail transit system is applied to construction of various cities, the urban rail transit is transformed from single-line operation to line network operation, the scale and complexity of the urban rail transit system are improved, and meanwhile challenges are provided for network management and development of the rail transit system. The rapid and accurate passenger flow prediction is not only the basis for scientifically making a driving plan, but also an important basis for adjusting an operation plan in real time, and is beneficial to traffic operation management to play a role more comprehensively and excellently. The Support Vector Machine (SVM) is used as a convex quadratic programming problem, and the obtained extreme value solution is guaranteed to be a global optimal solution. These characteristics make SVMs sufficient as an excellent data-based machine learning method. The SVM has outstanding performance in solving the problems of small samples, high-dimensional pattern recognition and the like, and can be applied and popularized to function fitting and other related machine learning researches. Different from the traditional machine learning algorithm, the SVM maps the original sample space to a high-dimensional feature space, and obtains the optimal linear classification surface in a new space. This mapping transformation, i.e. the non-linear transformation, is implemented using a suitable inner product function. The SVM successfully solves the local minimum value problem and the high-dimensional problem. And various learning algorithms such as a Bayes classifier, a radial basis function method, a multilayer perceptron network and the like can be realized by defining different inner product functions. The SVM uses a large interval factor to control the training process of machine learning, so that the SVM selects only the classification hyperplane with the maximum classification interval. The SVM algorithm has a complete theoretical basis, and has very good generalization performance in the application of some fields, so that the SVM algorithm has unsophisticated effects in the solution of classification, regression and density function estimation, and is successfully applied to the aspects of regression estimation, mode recognition and the like. Such as text classification, speech recognition, etc. SVMs were originally designed for classification problems and are widely used. In recent years, it has also shown very good performance in terms of regression problems.

The existing rail transit real-time passenger flow prediction method comprises the step that Roos J provides a dynamic Bayesian network method for predicting short-term passenger flow in a Paris railway network, and the method has a good processing effect on the condition of data insufficiency caused by system failure. The existing rail transit real-time passenger flow prediction method has the technical problems of low prediction accuracy and large calculation amount. Therefore, it is necessary to provide a rail transit real-time passenger flow prediction method with high prediction accuracy and small calculation amount.

Disclosure of Invention

The invention aims to solve the technical problems of low prediction accuracy and large calculation amount in the prior art. The rail transit real-time passenger flow prediction method has the characteristics of high prediction accuracy and small calculation amount.

In order to solve the technical problems, the technical scheme is as follows:

a rail transit real-time passenger flow prediction method, comprising:

(1) acquiring n pieces of historical data from an automatic fare collection system as original samples, and preprocessing the original samples to obtain preprocessed samples, wherein the preprocessed samples comprise time sequence vectors X and passenger flow vectors Y corresponding to the time sequence vectors X;

(2) establishing a short-time passenger flow prediction model based on a support vector regression machine according to the preprocessed samples in the step (1) and a kernel function and a fitting regression function, wherein the kernel function is a radial basis function;

(3) taking the time series vector X and the passenger flow vector Y corresponding to the time series vector X as input fitting prediction functions, inputting the input fitting prediction functions into the short-time passenger flow prediction model based on the support vector regression in the step (2), and predicting the passenger flow vector Y_n+1；

Wherein X ═ t₁,t₂,...,t_n}，Y＝{y₁,y₂,...,y_n}。

The working principle of the invention is as follows: the time series is a sequence of numbers ordered by time. The analysis process of the time series is to use the statistically observed time series data as a sample and build a model for predicting the occurrence of future events. The method comprises the following two steps: first, the continuity of event development is acknowledged; secondly, the randomness of the occurrence of the event is considered. The time series prediction mainly reflects three change rules mainly including periodic change, trend change and random change. The short-time passenger flow data has the change rule, and the prediction of the short-time passenger flow aims to help the rail transit operation safety operation and improve the service efficiency and the service quality. The passenger flow trend can be accurately predicted by establishing a high-precision prediction model. The passenger flow data has complexity and mutability, so that a linear prediction method is excluded, and a non-linear prediction method is selected.

Because the original samples are a large amount of data collected from the real world, and diversity, uncertainty and complexity exist between real production and actual life and scientific research, the collected original data are scattered, and the standard degree of the knowledge acquisition research conforming to the prediction algorithm is low. Therefore, the raw data must be processed and converted first before prediction is performed.

The method comprises the steps of collecting n pieces of historical data from an automatic fare collection system as original samples, preprocessing the original samples to obtain preprocessed samples, and establishing a short-time passenger flow prediction model based on a support vector regression machine according to a kernel function and a fitting regression function, wherein the kernel function is a radial basis function; and predicting future passenger flow according to a short-time passenger flow prediction model based on a support vector regression. The support vector regression model has only one type of sample points, and the optimal hyperplane sought is to minimize the total deviation value of all the sample points from the hyperplane. At this point, when the sample points are all contained between the two boundaries, finding the optimal regression hyperplane is equivalent to finding the maximum separation. The nonlinear support vector regression mainly uses a predetermined nonlinear mapping to map an input vector to a high-dimensional feature space, and performs linear regression in the high-dimensional feature space, so as to obtain the same effect as the original spatial nonlinear regression. In several kernel functions, a linear kernel function cannot process input values, the number of parameters of a high-dimensional kernel in a radial basis kernel function is less than that of a polynomial kernel function, but in the training process of a support vector regression machine, the training time required by the experiment of the polynomial kernel function is far longer than that of the radial basis kernel function, and when a Sigmoid kernel function is adopted, values in some parameters are wrong. Thus, the present invention employs a radial basis kernel function.

In the foregoing technical solution, for optimization, further, the establishing a short-time passenger flow prediction model based on a support vector regression includes:

(A) using the preprocessed samples as a given training set T { (x)₁,y₁),…,(x_n,y_n)}∈(R×y)ⁿ；

(B) According to the kernel function K (x, x'), the calculation precision is greater than 0 and the penalty function C is greater than 0, a convex quadratic programming target function is constructed and solved;

K(x_i,x_j)＝Φ(xi)·Φ(x_j)；

(C) calculating a decision function according to a quadratic programming target function and a nonlinear fitting function:

the nonlinear fitting function is:

the calculated decision function is:

(D) according to a decision function, taking n pieces of historical data as a training set, performing regression prediction on an n + 1-dimensional hyperplane, and calculating to obtain the short-time passenger flow prediction model as follows:

f(x+1)＝β₁f(x)+β₂f(x-1)+...+β_nf(x-n+1)；

where K (x, x') is the kernel function, x_i∈Rⁿ,y_i∈ y R, i 1 …, n, f (x +1) indicates the passenger flow at time x +1, β_iRepresents a weight coefficient, i ═ 1, 2.

Further, the radial basis function is a gaussian kernel function:

further, the passenger flow vector Y_n+1Including short-term traffic prediction and peak prediction.

Further, the peak prediction includes an early peak prediction and a late peak prediction.

Further, the preprocessing comprises key information extraction, data arrangement and minute-level passenger flow volume entering and exiting statistics.

Further, the key information extraction comprises extracting parameters of the card swiping time, the platform and the traffic card type 3 as key information.

Further, the statistical frequency of the minute-scale passenger flow volume statistics is not less than once every 5 minutes.

Support vector regression, i.e., support vector regression machines, is generally classified into linear regression and nonlinear regression. For linear regression, the sample data is estimated using a linear regression function. For nonlinear regression, data is mapped to a high-dimensional feature space through nonlinear mapping, linear regression is carried out in the space, the nonlinear regression adds a step of increasing dimensions, and the linear regression is carried out in the high-dimensional space to replace the nonlinear regression in a low-dimensional space, so that complex dot product operation in the high-dimensional space is omitted. The support vector regression algorithm is converted into the solution of the convex quadratic programming problem, and the scale of the solution is twice of that of the SVM classification problem under the same sample size.

Support vector regression mainly includes both linear and non-linear cases.

In a linear support vector regression machine, a training set is setFinding parameters in the regression function f (x, α) ═ ω · x + b using-insensitive loss function calculationAndnamely:

(ω·x_i+b)-y_i≤+ξ_i,i＝1,…,l

wherein, ξ_iAndis a relaxation variable, introduces Lagrange function

Wherein Lagrange multipliers satisfy

Let L pair b, w, ξ^(*)Is 0 with respect to α^(*)Maximum value of (a):

non-linear regression machineIn the first place, non-linear mapping is usedMapping the corresponding data to a high-dimensional feature space, and performing linear regression in the high-dimensional feature space. In the optimization process, only the characteristic space inner product operation is considered, so that the kernel function k (x, y) is used for replacingNon-linear regression can be achieved. The nonlinear regression is:

wherein,

find α^(*)Value of (1), mostα with values of 0 and not 0^(*)The corresponding sample is the support vector. The expression of f (x) is:

wherein, b is:

any support vector can calculate the b value.

The method provided by the invention is used for verifying the effectiveness of the prediction model of the support vector regression machine and comparing the fitting effect of the Gaussian kernel function and the RBF kernel function. The realization effect shows that the passenger flow prediction model based on the support vector regression has certain prediction capability, and the prediction conclusion is effective. Among the two kernel functions, the prediction effect of the Gaussian kernel function is obviously better than that of the RBF kernel function, and preferably, the Gaussian kernel function is selected as the optimal choice. The original sample is obtained from an Automatic Fare Collection (AFC) system in practical application, and the obtained passenger access information at any time is messy due to a slightly large experimental data volume. Therefore, an integration and key information extraction are required for a huge data set. In the selection of key information, the invention selects three kinds of key information from 7 kinds of labels, wherein the three kinds of key information comprise card swiping time, platform and traffic card type. The invention traverses all experimental data and arranges all contents of key information. The original data form is disordered, so that experimental prediction is not facilitated. Therefore, the data are arranged according to a certain rule, so that the subsequent experiment can be favorably carried out. The invention predicts the travel passenger flow of different stations, takes the station number in the station list arranged in the previous step as a classification label, and classifies the entrance and exit records of the rail transit network in a certain time. The original data of the invention contains a plurality of information, and each piece of information only records the access state of a certain passenger at a certain moment. If a prediction model is to be built for passenger flow prediction, the original data must be counted. Therefore, the number of the inbound people and the number of the outbound people of the station are counted every 5 minutes by traversing each sorted station data file.

The invention has the beneficial effects that:

the method has the advantages that the first effect is that the prediction accuracy is improved;

the second effect is that the prediction efficiency is improved;

and the third effect is that the data is preprocessed, so that the diversity, uncertainty and complexity of prediction are reduced.

Drawings

The invention is further illustrated with reference to the following figures and examples.

FIG. 1 is a schematic diagram of a data preprocessing flow.

Fig. 2, No. 0321 station arrival prediction support vector regression model.

Fig. 3, No. 0321 station outbound prediction support vector regression model.

Figure 4, No. 0212 station arrival prediction support vector regression model.

Figure 5, No. 0212 station outbound predicted support vector regression model.

The regression model of station entry prediction support vector of 0315 in fig. 6.

The model of the predictive support vector regression for station number 0315 in fig. 7.

Fig. 8, No. 0613 platform early peak prediction support vector regression model.

Fig. 9, 0210 station early peak prediction support vector regression model.

The model of model vector regression is supported by station late peak prediction of fig. 10 and 0613.

Figure 11, 0210 station late peak prediction support vector regression model.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail with reference to the following embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

Example 1

The embodiment provides a rail transit real-time passenger flow prediction method, which comprises the following steps:

Wherein X ═ t₁,t₂,...,t_n}，Y＝{y₁,y₂,...,y_n}。

Specifically, the establishing of the short-time passenger flow prediction model based on the support vector regression includes:

K(x_i,x_j)＝Φ(x_i)·Φ(x_j)；

the nonlinear fitting function is:

the calculated decision function is:

f(x+1)＝β₁f(x)+β₂f(x-1)+...+β_nf(x-n+1)；

Preferably, further, the radial basis function is a gaussian kernel function:

in particular, the passenger flow vector Y_n+1Including short-term traffic prediction and peak prediction.

More specifically, the peak forecast includes an early peak forecast and a late peak forecast.

Because the original samples are a large amount of data collected from the real world, and diversity, uncertainty and complexity exist between real production and actual life and scientific research, the collected original data are scattered, and the degree of conforming to a prediction algorithm to perform knowledge acquisition research is low.

Preferably, as shown in fig. 1, the preprocessing includes key information extraction, data sorting and minute-level incoming and outgoing passenger flow statistics. Specifically, the key information extraction in this embodiment includes extracting parameters of the card swiping time, the platform and the traffic card type 3 as key information.

The original sample is obtained from an automatic ticket selling and checking system in practical application, and due to the large data volume, the obtained passenger entering and exiting information at any time is messy. Therefore, an integration and key information extraction are required for huge data sets. In the selection of the key information, three tags with more research value are selected from the 7 tags as the key information, which include the card-swiping time, the platform and the traffic card type. In the embodiment, all experimental data are traversed in the experimental process, and all contents of the key information are sorted. And (4) finishing results: table 1 is a data sample keyword time list, table 2 is a data sample keyword platform list, and table 3 is a data sample keyword card class number list.

Time of day	Time of day
		20140101	20140117
20140102	20140118
		20140103	20140119
20140104	20140120
		20140105	20140121
20140106	20140122
		20140107	20140123
20140108	20140124
		20140109	20140125
20140110	20140126
		20140111	20140127
20140112	20140128
		20140113	20140129
20140114	20140130
		20140115	20140131
20140116

TABLE 1

Station number	Station number	Station number	Station number
				0102	0201	0301	0609
0103	0202	0302	0610
				0104	0203	0303	0611
0105	0204	0304	0613
				0106	0205	0305	0614
…	…	…	…
				0120	0215	0336	0623
0121	0216	0337	0625
				0122	0217	0338	0626
0123	0218	0339	0628

TABLE 2

Card number	Card type	Card number	Card type
				00	Common card for livable	05	Yijuyue Ticket
14	Free card for living in	77	Free counting card for livable living
				15	Livable employee's card	82	Official ticket
20	Love preferential card	88	Employee's ticket
				03	Card for student living in livable house	89	One-way commemorative ticket
44	Rail police on-duty card	94	Track order ticket
				48	Track service card	98	One-way ticket

TABLE 3

The validation data collection times shown in table 1 ranged from 1 month 1 day 2014 to 1 month 31 day 2014. As can be seen from table 2, the first two digits of the stations of the statistical data represent the names of the rail transit lines, and the last two digits represent the station numbers and increase from beginning to end. Wherein the missing station number indicates that the station on the line has not been opened. The traffic card type statistics are shown in table 3, and the card type name corresponding to the number is also correspondingly shown.

The original data form is disordered, so that experimental prediction is not facilitated. Therefore, the data are arranged according to a certain rule, so that the subsequent experiment can be favorably carried out. And planning to predict the travel passenger flow of different stations, and taking the station numbers in the station list sorted in the previous step as classification labels to classify the access records of the rail transit network in one month. Each site file contains records of the flow of passengers and passengers for that site in month 1 of 2014. The original data is more, and each piece of information only records the access state of a certain passenger at a certain moment. If a prediction model is built for passenger flow prediction, the calculation amount can be reduced by counting the original data. Thus, each consolidated station data file is traversed.

Preferably, the number of inbound people and the number of outbound people for the site are counted every 5 minutes, and the statistics are shown in Table 4.

Counting the start time	Statistical cut-off time	The number of people entering the room	Number of people going out
				20140112-070000	20140112-070500	29	63
20140112-070500	20140112-071000	18	67
				20140112-071000	20140112-071500	23	25
20140112-071500	20140112-072000	18	86
				20140112-072000	20140112-072500	21	143
20140112-072500	20140112-073000	24	112
				20140112-073000	20140112-073500	25	36
20140112-073500	20140112-074000	43	142
				20140112-074000	20140112-074500	30	135

TABLE 4

The prediction result of the embodiment is as follows:

short-time passenger flow prediction: selecting a station No. 0321, a station No. 0212 and a station No. 0315 as prediction objects, training the inbound/outbound data of the station from 15 hours to 21 hours in the period from 1 month 6 to 1 month 21 in 2014, predicting the inbound/outbound passenger flow in the same period on 22 days in 2014, and counting the training data once every 15 minutes. The experimental results are shown in FIGS. 2-7. Wherein, fig. 2 and fig. 3 are the predicted support vector regression model for station No. 0321 entering/exiting respectively; fig. 4 and 5 are schematic diagrams of predicted support vector regression models for station 0212 entering/exiting respectively; fig. 6 and 7 show the model of the predictive support vector regression for station number 0315.

Wherein the abscissa is a scale every 5 minutes during the time period, each of the projections represents the traffic situation during business hours of the day, and the last projection is the predicted data.

Peak prediction:

early peak prediction: selecting the platform No. 0613 and the platform No. 0210 as prediction objects, training the arrival data of the station from 6 hours to 30 hours to 9 hours every day from 1 month and 6 days to 1 month and 21 days in 2014, predicting the arrival passenger flow of the station in the same period of 22 days in 2014, and counting the training data once every 15 minutes. The experimental results are shown in fig. 8 and 9, where fig. 8 is the model of the early peak prediction support vector regression for station 0613 and fig. 9 is the model of the early peak prediction support vector regression for station 0210. Wherein the abscissa is a scale every 20 minutes during the time period.

And (3) late peak prediction: selecting the platform No. 0613 and the platform No. 0210 as prediction objects, training the arrival data of the station from 17 hours to 20 hours every day from 1 month 6 days to 1 month 21 days in 2014, predicting the arrival passenger flow in the same period of 22 days in 2014, and counting the training data once every 15 minutes. The experimental results are shown in fig. 10 and fig. 11, where fig. 10 is a model of model 0613 for model 0210 for model.

Wherein the abscissa is a scale every 20 minutes during the time period.

Although the illustrative embodiments of the present invention have been described above to enable those skilled in the art to understand the present invention, the present invention is not limited to the scope of the embodiments, and it is apparent to those skilled in the art that all the inventive concepts using the present invention are protected as long as they can be changed within the spirit and scope of the present invention as defined and defined by the appended claims.

Claims

1. A rail transit real-time passenger flow prediction method is characterized by comprising the following steps: the method comprises the following steps:

Wherein X ═ t₁,t₂,...,t_n}，Y＝{y₁,y₂,...,y_n}。

2. The rail transit real-time passenger flow prediction method according to claim 1, characterized in that: the establishing of the short-time passenger flow prediction model based on the support vector regression comprises the following steps:

<mfenced open = "" close = ""> <mtable> <mtr> <mtd> <mrow> <mi>W</mi> <mrow> <mo>(</mo> <mrow> <mi>a</mi> <mo>,</mo> <msup> <mi>a</mi> <mo>*</mo> </msup> </mrow> <mo>)</mo> </mrow> <mo>=</mo> <mo>-</mo> <mfrac> <mn>1</mn> <mn>2</mn> </mfrac> <munderover> <mi>&Sigma;</mi> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> <mo>,</mo> <mi>j</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <mrow> <mo>(</mo> <mrow> <msub> <mi>a</mi> <mi>i</mi> </msub> <mo>-</mo> <msubsup> <mi>a</mi> <mi>i</mi> <mo>*</mo> </msubsup> </mrow> <mo>)</mo> </mrow> <mrow> <mo>(</mo> <mrow> <msub> <mi>a</mi> <mi>j</mi> </msub> <mo>-</mo> <msubsup> <mi>a</mi> <mi>j</mi> <mo>*</mo> </msubsup> </mrow> <mo>)</mo> </mrow> <mo>&CenterDot;</mo> <mi>K</mi> <mrow> <mo>(</mo> <mrow> <mi>x</mi> <mo>,</mo> <msub> <mi>x</mi> <mi>i</mi> </msub> </mrow> <mo>)</mo> </mrow> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <mo>+</mo> <munderover> <mi>&Sigma;</mi> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> <mo>,</mo> <mi>j</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <mrow> <mo>(</mo> <mrow> <msub> <mi>a</mi> <mi>i</mi> </msub> <mo>-</mo> <msubsup> <mi>a</mi> <mi>i</mi> <mo>*</mo> </msubsup> </mrow> <mo>)</mo> </mrow> <msub> <mi>y</mi> <mi>i</mi> </msub> <mo>-</mo> <munderover> <mi>&Sigma;</mi> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> <mo>,</mo> <mi>j</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <mrow> <mo>(</mo> <mrow> <msub> <mi>a</mi> <mi>i</mi> </msub> <mo>+</mo> <msubsup> <mi>a</mi> <mi>i</mi> <mo>*</mo> </msubsup> </mrow> <mo>)</mo> </mrow> <mi>&epsiv;</mi> </mrow> </mtd> </mtr> </mtable> </mfenced>

K(x_i,x_j)＝Φ(x_i)·Φ(x_j)；

the nonlinear fitting function is:

<mfenced open = "" close = ""> <mtable> <mtr> <mtd> <mrow> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>)</mo> </mrow> <mo>=</mo> <mi>&omega;</mi> <mo>&CenterDot;</mo> <mi>&Phi;</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>)</mo> </mrow> <mo>+</mo> <mi>b</mi> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <mo>=</mo> <munderover> <mo>&Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <mrow> <mo>(</mo> <msub> <mi>a</mi> <mi>i</mi> </msub> <mo>-</mo> <msubsup> <mi>a</mi> <mi>i</mi> <mo>*</mo> </msubsup> <mo>)</mo> </mrow> <mi>K</mi> <mo>(</mo> <mrow> <mi>x</mi> <mo>,</mo> <msub> <mi>x</mi> <mi>i</mi> </msub> </mrow> <mo>)</mo> <mo>+</mo> <mi>b</mi> </mrow> </mtd> </mtr> </mtable> </mfenced>

the calculated decision function is:

<mrow> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>)</mo> </mrow> <mo>=</mo> <munderover> <mo>&Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <mrow> <mo>(</mo> <msubsup> <mover> <mi>&alpha;</mi> <mo>&OverBar;</mo> </mover> <mi>i</mi> <mo>*</mo> </msubsup> <mo>-</mo> <msub> <mover> <mi>&alpha;</mi> <mo>&OverBar;</mo> </mover> <mi>i</mi> </msub> <mo>)</mo> </mrow> <mi>K</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>)</mo> </mrow> <mo>+</mo> <mover> <mi>b</mi> <mo>&OverBar;</mo> </mover> <mo>;</mo> </mrow>

f(x+1)＝β₁f(x)+β₂f(x-1)+...+β_nf(x-n+1)；

3. The rail transit real-time passenger flow prediction method according to claim 2, characterized in that: the radial basis function is a gaussian kernel function:

<mrow> <mi>k</mi> <mrow> <mo>(</mo> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>,</mo> <msub> <mi>x</mi> <mi>j</mi> </msub> <mo>)</mo> </mrow> <mo>=</mo> <mi>exp</mi> <mrow> <mo>(</mo> <mo>-</mo> <mfrac> <mrow> <mo>|</mo> <mo>|</mo> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>-</mo> <msub> <mi>x</mi> <mi>j</mi> </msub> <mo>|</mo> <msup> <mo>|</mo> <mn>2</mn> </msup> </mrow> <mrow> <mn>2</mn> <msup> <mi>&sigma;</mi> <mn>2</mn> </msup> </mrow> </mfrac> <mo>)</mo> </mrow> <mo>.</mo> </mrow>

4. the rail transit real-time passenger flow prediction method according to claim 1, characterized in that: the passenger flow vector Y_n+1Including short-term traffic prediction and peak prediction.

5. The rail transit real-time passenger flow prediction method according to claim 4, characterized in that: the peak prediction includes an early peak prediction and a late peak prediction.

6. The rail transit real-time passenger flow prediction method according to claim 1, characterized in that: the preprocessing comprises key information extraction, data arrangement and minute-level passenger flow volume entering and exiting statistics.

7. The rail transit real-time passenger flow prediction method of claim 6, characterized in that: the key information extraction comprises extracting parameters in the card swiping time, the platform and the traffic card type 3 as key information.

8. The rail transit real-time passenger flow prediction method of claim 6, characterized in that: the statistical frequency of the minute-level passenger flow volume is not less than 5 minutes.