CN112489422A - Traffic flow prediction method based on algorithm model analysis - Google Patents
Traffic flow prediction method based on algorithm model analysis Download PDFInfo
- Publication number
- CN112489422A CN112489422A CN202011304652.6A CN202011304652A CN112489422A CN 112489422 A CN112489422 A CN 112489422A CN 202011304652 A CN202011304652 A CN 202011304652A CN 112489422 A CN112489422 A CN 112489422A
- Authority
- CN
- China
- Prior art keywords
- traffic flow
- data
- sequence
- model
- detected
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
- G08G1/0125—Traffic data processing
- G08G1/0129—Traffic data processing for creating historical data or processing based on historical data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
- G08G1/0137—Measuring and analyzing of parameters relative to traffic conditions for specific applications
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/065—Traffic control systems for road vehicles by counting the vehicles in a section of the road or in a parking area, i.e. comparing incoming count with outgoing count
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses a traffic flow prediction method based on algorithm model analysis, which is different from the existing mode of carrying out data analysis on single influence factors. Before the ARIMA model is used for the data sequence, the numerical value obtained by the LSTM model is subjected to iterative optimization through a loss function of data obtained by the mse function, the data viscosity is increased, the optimal time sequence to be measured is obtained, the accuracy of data prediction is improved, and a new thought is provided for big data processing.
Description
Technical Field
The application belongs to the technical field of intelligent public traffic data processing, and particularly relates to a traffic flow prediction method based on algorithm model analysis.
Background
Along with the rapid development of economic society and the continuous acceleration of urbanization, the quantity of motor vehicles in China is increasing year by year. The traffic flow prediction has high guiding significance for public transport travel, scheduling and other works. However, in the existing traffic flow prediction mode, simple prediction is basically performed according to the influence factors such as time points, weather, holidays and the like, and the considered influence factors are single, so that the final traffic flow prediction result is not accurate enough, and the data referential is greatly reduced.
The current prediction method based on the time sequence has few required parameters and wide prediction application range, and has a certain achievement on short-term traffic flow prediction like the current ARIMA model, SARIMA model and VAR model. However, when the prediction is performed by using the models, only the models themselves are usually looked at, and no good processing mode exists for data input into the models at present, which causes that even if the accuracy of the models in the prediction is high, the prediction result is inaccurate due to inaccuracy of data affected by multiple factors.
Disclosure of Invention
The application aims to provide a traffic flow prediction method based on algorithm model analysis, and aims to solve the problem that the accuracy of traffic flow prediction in the prior art is not high.
In order to achieve the purpose, the technical scheme adopted by the application is as follows:
a traffic flow prediction method based on algorithm model analysis is used for traffic flow analysis early warning in public intelligent transportation and comprises the following steps:
step 1, obtaining historical traffic flow data, performing principal component analysis and classification on the historical traffic flow data based on four influence factors, namely weather, period, season and intersection environment, obtaining initial data to be detected including four different groups, and calculating the proportion of data quantity in each group in the initial data to be detected;
step 2, dividing the initial data to be measured into a plurality of sets again according to a partition coefficient, numbering all historical traffic flow data in each set respectively, forming a one-dimensional sequence to be measured from all the historical traffic flow data in each set according to the numbering, inputting all the one-dimensional sequences to be measured into an LSTM model which is iteratively optimized based on a mse function respectively to obtain a plurality of corresponding time sequences to be measured, and calculating the partition coefficient according to the ratio of each group in the initial data to be measured;
step 3, judging the stationarity of the time sequence to be detected according to the autocorrelation coefficient and the partial autocorrelation coefficient of the historical traffic flow data, and if the time sequence to be detected is a stable sequence, carrying out white noise process treatment; otherwise, judging whether the time sequence is a stable sequence or not again after the first-order difference processing is carried out, and carrying out white noise process processing if the time sequence to be detected after the first-order difference processing is the stable sequence; otherwise, carrying out white noise process after carrying out second-order differential processing;
step 4, based on the time sequence to be detected after white noise process processing, utilizing the ARIMA model to predict the traffic flow, and if the confidence level output by the ARIMA model is less than a confidence threshold, discarding the prediction result output by the current ARIMA model; otherwise, the prediction result output by the ARIMA model is kept;
and 5, executing the step 4 aiming at each time sequence to be detected after the white noise process is processed to obtain a plurality of prediction results, and fusing the plurality of prediction results to obtain a traffic flow prediction value of the intersection corresponding to the historical traffic flow data.
Several alternatives are provided below, but not as an additional limitation to the above general solution, but merely as a further addition or preference, each alternative being combinable individually for the above general solution or among several alternatives without technical or logical contradictions.
Preferably, the step of respectively inputting all the one-dimensional time sequences to be tested into an LSTM model that is iteratively optimized based on an mse function to obtain a plurality of corresponding time sequences to be tested includes:
inputting a one-dimensional sequence to be detected into an LSTM model;
calculating a loss function through a mean square error mse function based on a current one-dimensional sequence to be measured and a time sequence to be measured output by an LSTM model;
if the loss value calculated according to the loss function is larger than the preset loss threshold value, adjusting the parameters of the LSTM model and inputting the one-dimensional sequence to be measured into the LSTM model after the parameters are adjusted again for processing; otherwise, taking the time sequence to be tested output by the current LSTM model as the final time sequence to be tested;
and taking a new one-dimensional sequence to be detected and repeating the steps until the processing of all the one-dimensional sequences to be detected is completed.
Preferably, the step 2 further includes, before the initial data to be measured is re-divided into a plurality of sets according to the segmentation coefficients, performing smoothing on the initial data to be measured, where the smoothing includes:
calculating the average traffic flow of each group according to all historical traffic flow data in each group;
and removing the historical traffic flow data of which the absolute value of the difference value between the historical traffic flow data and the average traffic flow data in each group is larger than the preset multiple of the average traffic flow.
Preferably, the white noise process processing includes:
defining an observed value z of traffic flow datat：
zt＝λ1zt+1+λ2zt-2+λ3zt-3+…+pzt-p+vt
In the formula, observed value ztI.e. the autoregressive model of the number of hysteretic variables at time t, lambdapIs a regression parameter between historical traffic flow data and influencing factors, p is the number of lagging variables at peak time of traffic flow, vtRepresenting a noise value of the white noise processing process of the historical traffic flow data at the time t;
and substituting each historical traffic flow data in the time sequence to be detected into the autoregressive model to obtain a corresponding observed value, and finishing white noise process processing.
Preferably, the traffic flow prediction method based on algorithmic model analysis further includes calculating a traffic flow data change trend, where calculating the traffic flow data change trend includes:
constructing an increase ratio index, wherein the expression is as follows:
in the formula, ziIndicates the predicted value of the traffic flow obtained by the current prediction, zi-1And the traffic flow data of the previous time period i-1 relative to the time period i where the current traffic flow predicted value is located.
Compared with the prior art, the traffic flow prediction method based on the algorithm model analysis has the following beneficial effects:
(1) different from the existing mode of carrying out data analysis on single influence factor, the method carries out principal component analysis, division and classification on various influence factors (such as weather, period, season, intersection environment and the like) related to traffic flow change, and can enable a predicted value to be more representative;
(2) before the ARIMA model is used for the data sequence, the numerical value obtained by the LSTM model is subjected to iterative optimization through a loss function of data obtained by the mse function, the data viscosity is increased, the optimal time sequence to be measured is obtained, the accuracy of data prediction is improved, and a new thought is provided for big data processing;
(3) relevant processing such as dimension reduction segmentation and the like is already carried out when the initial sequence is learned and predicted, and the running speed of subsequent algorithm calculation can be improved.
Drawings
Fig. 1 is a flowchart of a traffic flow prediction method based on algorithmic model analysis according to the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in the description of the present application herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application.
The real-time and accurate road short-time traffic prediction is the core content and important basis of an intelligent traffic system, and is beneficial to a traffic manager to implement dynamic traffic guidance, ensure traffic control and safety and realize advanced traffic management. The traffic flow prediction method based on the algorithm model analysis can comprehensively consider the influence of various influence factors on traffic flow prediction, obtain a traffic flow prediction result with higher accuracy, lay a foundation for the operation of an intelligent traffic system, and have higher application value in the aspects of traffic resource scheduling realized based on traffic flow prediction values, traffic transportation safety guarantee and the like.
The method is particularly applied to the traffic flow prediction of the expressway toll station, can realize the dynamic configuration of expressway toll personnel based on the traffic flow prediction value, and has important reference significance for traffic departments to master the traffic flow condition of the expressway and implement traffic flow induction, diversion, information release and the like.
As shown in fig. 1, the traffic flow prediction method based on algorithmic model analysis according to the present embodiment includes the following steps:
step 1, obtaining historical traffic flow data, performing principal component analysis and classification on the historical traffic flow data based on four influence factors, namely weather, period, season and intersection environment, obtaining initial data to be detected including four different groups, and calculating the proportion of data quantity in each group in the initial data to be detected.
It should be noted that the historical traffic data may be data from the same source (i.e., acquired by the same platform), or may be data from multiple sources, and the source of the data is not limited in this embodiment. Similarly, however, regardless of the source of the historical traffic data, each historical traffic data is tagged with a corresponding tag to identify the weather, period, season, and intersection environment to which the historical traffic data corresponds.
It is easily understood that setting the tags in order to identify the traffic data at different time intervals is a conventional means in the technical field of traffic data processing, and how the tags are set is not described here. The content of the label is also set in various ways according to practical application, for example, the weather can include rainy days, sunny days and snowy days, and the rainy days can be further refined into big rainy days, small rainy days and the like, the period refers to a calculation period which is a year, month, day or week, a quarter, a specific time interval point and the like, and the intersection environment can be divided into a human-shaped intersection, a three-way intersection, an intersection, whether to put in a signal lamp or not and the like. However, no matter how the label content is set, each piece of historical traffic data can be divided into groups closest to a certain influence factor, and the specific division process is an execution step of a principal component analysis method.
When historical traffic flow data are obtained, selecting according to set weather, period, season and intersection environment as screening conditions to obtain an intersection to be predicted and a prediction result of the next period predicted under the set period.
After the classification is performed in this embodiment, each group includes different data amounts, and in order to better fuse data in each group subsequently, a proportion of the data amount in each group to a total data amount (i.e., initial data to be measured) is calculated, so as to obtain a percentage model. Since the acquired historical traffic data includes a plurality of pieces of historical traffic data, the data amount here refers to the number of pieces of historical traffic data in each group. And 2, dividing the initial data to be measured into a plurality of sets again according to a division coefficient, numbering all historical traffic data in each set respectively, forming a one-dimensional sequence to be measured by all the historical traffic data in each set according to the numbering, inputting all the one-dimensional sequence to be measured into an LSTM (long-short term memory) model which is iteratively optimized based on a mse function respectively to obtain a plurality of corresponding time sequences to be measured, and calculating the division coefficient according to the ratio of each group in the initial data to be measured.
In order to make the division of the set more fit with the originally acquired historical traffic data, in this embodiment, the division coefficient is determined based on the ratio calculated in step 1, the ratio may be directly re-divided as a division ratio, or a normalized mean value of each ratio may be re-divided as a division coefficient, and this embodiment does not limit a specific division coefficient determination manner, and may be adjusted according to actual operation.
In this embodiment, the initial data to be measured classified in step 1 is re-divided, that is, the data in all the groups are re-mixed, and the data is re-divided based on the mixed disorderly sequence, so that each set contains data of each influence factor, and it is avoided that the final prediction result is inaccurate due to too single historical data.
And the LSTM model is utilized to process the one-dimensional time sequence to be detected and output the time sequence to be detected with stronger time sequence. Generally, an LSTM model is used for training and optimizing the LSTM model in advance to obtain an optimal LSTM model for training data, the complexity of influencing factors in historical traffic data is considered, the LSTM model is not trained and optimized in advance, the LSTM model which is iteratively optimized based on a mse function is adopted for output, and the following modes are adopted when a time sequence to be measured is obtained in the specific embodiment:
1) inputting a one-dimensional sequence to be detected into an LSTM model;
2) calculating a loss function through a mean square error mse function based on a current one-dimensional sequence to be measured and a time sequence to be measured output by an LSTM model;
3) if the loss value calculated according to the loss function is larger than the preset loss threshold value, adjusting parameters of the LSTM model (including updating the weight of the gradient of each layer in each neuron) and inputting the one-dimensional sequence to be measured into the LSTM model after the parameters are adjusted again for processing, namely, executing the step 2 again after the parameters of the LSTM model are updated); otherwise, taking the time sequence to be tested output by the current LSTM model as the final time sequence to be tested;
4) and taking a new one-dimensional sequence to be detected and repeating the steps 2) and 3) until the processing of all the one-dimensional sequences to be detected is completed.
The LSTM model has default parameters before processing, the first one-dimensional sequence to be measured is processed based on the default parameters for the first time, and other one-dimensional sequences to be measured can be processed based on the LSTM model with the default parameters or can be processed based on the finally updated parameters of the last one-dimensional sequence to be measured. In the embodiment, when the LSTM model is used, each one-dimensional to-be-measured sequence is used as training data to optimize the LSTM model based on the mse function, and the optimized LSTM model is used to output a final to-be-measured time sequence.
The calculation formula of the mean square error mse function is as follows:
in the formula, n is the average traffic flow value of the current set (i.e. one-dimensional sequence to be measured), m is the data volume of the current set, i is the ith number in the current set, and ω isiWeight, y, corresponding to the ith number in the current setiFor the ith numbered historical traffic data in the current set,
output of and y for LSTM modeliCorresponding predicted traffic flow data.
And (3) obtaining a loss function of data by adopting the mse function to carry out iterative optimization, calculating the gradient of each layer in each neuron to carry out weight updating, and carrying out machine learning on different time point data through a large number of training time sequences to obtain influence factors related to traffic flow and weight ratio between the time points. And the result is zoomed to obtain a new time sequence for measuring the traffic flow.
For the traffic flow, the variation range is usually small, and the data with large variation range is usually not representative, so in another embodiment, to avoid the influence of the value with too large or too small variation range on the prediction result, in this embodiment, the data in each group in the initial data to be measured is gently processed, which specifically includes:
calculating the average traffic flow of each group according to all historical traffic flow data in each group;
and removing the historical traffic flow data of which the absolute value of the difference value with the average traffic flow in each group is larger than the preset multiple of the average traffic flow.
That is, if
Then the historical traffic data x needs to be comparedjRemoving, wherein
Is xjAnd b is a preset multiple of the average traffic flow of all historical traffic flow data of the group. The value of b is usually within 5% to ensure a better data smoothing effect.
Step 3, judging the stationarity of the time sequence to be detected according to the autocorrelation coefficient and the partial autocorrelation coefficient of the historical traffic flow data, and if the time sequence to be detected is a stable sequence, carrying out white noise process treatment; otherwise, judging whether the time sequence is a stable sequence or not again after the first-order difference processing is carried out, and carrying out white noise process processing if the time sequence to be detected after the first-order difference processing is the stable sequence; otherwise, carrying out white noise process after carrying out second-order difference process.
And when judging whether the time sequence to be detected is a stable sequence, substituting the judged time sequence to be detected into an autocorrelation and partial correlation coefficient calculation formula, and judging according to the characteristics that the stability of the stable sequence is mainly reflected in the aspects of unchanged mean value, limited variance and weak other limitations. It should be noted that, determining the stationarity of the sequence is a conventional process applied to the time sequence, and is not described herein again.
The subsequent prediction processing is carried out on the basis of the stationary sequence because in the stationary sequence, the (x) is often1,…,xn) And xn+1Not independent, it is possible to predict future times using historical samples.
The specific steps of performing white noise process processing in this embodiment are as follows:
defining an observed value z of traffic flow datat：
zt＝λ1zt-1+λ2zt-2+λ3zt-3+…+λpzt-p+vt
In the formula, observed value ztI.e. the autoregressive model of the number of hysteretic variables at time t, lambdapIs a regression parameter between historical traffic flow data and influencing factors, p is the number of lagging variables at peak time of traffic flow, vtRepresenting a noise value of the white noise processing process of the historical traffic flow data at the time t;
and substituting each historical traffic flow data in the time sequence to be detected into the autoregressive model to obtain a corresponding observed value, and finishing white noise process processing. The white noise process converts the common time series data into a data format that can be identified by an ARIMA Model (Autoregressive Integrated Moving Average Model) so as to facilitate prediction, and is a basic processing step for prediction by using the ARIMA Model.
Step 4, based on the time sequence to be detected after white noise process processing, utilizing the ARIMA model to predict the traffic flow, and if the confidence level output by the ARIMA model is less than a confidence threshold, discarding the prediction result output by the current ARIMA model; otherwise, the prediction result output by the ARIMA model is kept.
When the vehicle flow is predicted, the prediction results with unsatisfactory prediction results are discarded, only the prediction results with high confidence level are reserved, and the final prediction results can be further screened to improve the final prediction value. The normal confidence threshold is set to be 95%, so that the larger calculation pressure caused by the cycle number process is avoided, and the prediction result with higher accuracy can be ensured.
And 5, executing the step 4 aiming at each time sequence to be detected after the white noise process is processed to obtain a plurality of prediction results, and fusing the plurality of prediction results to obtain a traffic flow prediction value of the intersection corresponding to the historical traffic flow data.
If no prediction result is reserved after the step 4 is executed on each time sequence to be tested, the set division representativeness is low, and therefore the step 2 is returned to for re-division and execution; if only one prediction result is reserved, taking the prediction result as a traffic flow prediction value of the current intersection (the intersection is the intersection corresponding to the historical traffic flow data acquired in the step 1); if a plurality of prediction results are reserved, the plurality of prediction results are fused, and the fusion may be to select a random one of the plurality of prediction results, or to take the average value and the median value of the plurality of prediction results as the final traffic flow prediction result.
In practical applications, since the confidence levels of all the predicted results do not reach the confidence threshold value, a plurality of predicted results are generally considered, and a manner of obtaining the final predicted value of the traffic flow from the plurality of predicted results may be a plurality of manners, and a preferable manner is averaging.
The predicted traffic flow predicted value is a traffic flow predicted value for a next time period, where the next time period is determined according to preset starting predicted time and a time interval, the time interval is the period set in step 1, and the starting predicted time is usually the current time, or may be the latest time in the historical traffic flow data.
In one embodiment, in order to increase the situation analysis of the change rate of the predicted value, the change trend of the traffic flow of the predicted value under the algorithm analysis under the influence factors of weather, period, season and intersection environment is effectively further judged. The embodiment introduces a gain index, and the expression of the gain index is as follows:
in the formula, ziIndicates the predicted value of the traffic flow obtained by the current prediction, zi-1And the traffic flow data of the previous time period i-1 relative to the time period i where the current traffic flow predicted value is located.
The change rate of the traffic flow value in the time period interval can be obtained through the importing calculation, and the traffic flow change trend is judged so as to carry out prospective judgment on the traffic flow.
Compared with the traditional prediction model, the traffic flow obtained by the method has the advantages that the filtered initial data to be tested of the traffic flow are divided into a plurality of sets by the aid of the preset division coefficient, the processed data sets are divided into one-dimensional sequences to be tested, the sets are numbered, the processed data sets are combined with the LSTM model, loss functions of the data are obtained by means of the mse function, iterative optimization is carried out, the finally confirmed time sequence to be tested of the traffic flow is obtained, the obtained detection data are more representative and professional, and accuracy of traffic flow prediction results can be effectively improved.
It should be understood that, although the steps in the flowchart of fig. 1 are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least a portion of the steps in fig. 1 may include multiple sub-steps or multiple stages that are not necessarily performed at the same time, but may be performed at different times, and the order of performance of the sub-steps or stages is not necessarily sequential, but may be performed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.
The technical features of the embodiments described above may be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the embodiments described above are not described, but should be considered as being within the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (5)
1. A traffic flow prediction method based on algorithm model analysis is used for traffic flow analysis early warning in public intelligent transportation and is characterized in that the traffic flow prediction method based on algorithm model analysis comprises the following steps:
step 1, obtaining historical traffic flow data, performing principal component analysis and classification on the historical traffic flow data based on four influence factors, namely weather, period, season and intersection environment, obtaining initial data to be detected including four different groups, and calculating the proportion of data quantity in each group in the initial data to be detected;
step 2, dividing the initial data to be measured into a plurality of sets again according to a partition coefficient, numbering all historical traffic flow data in each set respectively, forming a one-dimensional sequence to be measured from all the historical traffic flow data in each set according to the numbering, inputting all the one-dimensional sequences to be measured into an LSTM model which is iteratively optimized based on a mse function respectively to obtain a plurality of corresponding time sequences to be measured, and calculating the partition coefficient according to the ratio of each group in the initial data to be measured;
step 3, judging the stationarity of the time sequence to be detected according to the autocorrelation coefficient and the partial autocorrelation coefficient of the historical traffic flow data, and if the time sequence to be detected is a stable sequence, carrying out white noise process treatment; otherwise, judging whether the time sequence is a stable sequence or not again after the first-order difference processing is carried out, and carrying out white noise process processing if the time sequence to be detected after the first-order difference processing is the stable sequence; otherwise, carrying out white noise process after carrying out second-order differential processing;
step 4, based on the time sequence to be detected after white noise process processing, utilizing the ARIMA model to predict the traffic flow, and if the confidence level output by the ARIMA model is less than a confidence threshold, discarding the prediction result output by the current ARIMA model; otherwise, the prediction result output by the ARIMA model is kept;
and 5, executing the step 4 aiming at each time sequence to be detected after the white noise process is processed to obtain a plurality of prediction results, and fusing the plurality of prediction results to obtain a traffic flow prediction value of the intersection corresponding to the historical traffic flow data.
2. The traffic flow prediction method based on algorithmic model analysis according to claim 1, wherein said inputting all one-dimensional sequences to be tested into an LSTM model that is iteratively optimized based on an mse function to obtain a plurality of corresponding time sequences to be tested comprises:
inputting a one-dimensional sequence to be detected into an LSTM model;
calculating a loss function through a mean square error mse function based on a current one-dimensional sequence to be measured and a time sequence to be measured output by an LSTM model;
if the loss value calculated according to the loss function is larger than the preset loss threshold value, adjusting the parameters of the LSTM model and inputting the one-dimensional sequence to be measured into the LSTM model after the parameters are adjusted again for processing; otherwise, taking the time sequence to be tested output by the current LSTM model as the final time sequence to be tested;
and taking a new one-dimensional sequence to be detected and repeating the steps until the processing of all the one-dimensional sequences to be detected is completed.
3. The traffic flow prediction method based on algorithmic model analysis as defined in claim 1, wherein step 2 further comprises smoothing the initial data to be measured before subdividing the initial data to be measured into a plurality of sets according to the segmentation coefficients, the smoothing comprising:
calculating the average traffic flow of each group according to all historical traffic flow data in each group;
and removing the historical traffic flow data of which the absolute value of the difference value between the historical traffic flow data and the average traffic flow data in each group is larger than the preset multiple of the average traffic flow.
4. The method of traffic prediction based on algorithmic model analysis as defined in claim 1, wherein the white noise process comprises:
defining an observed value z of traffic flow datat：
In the formula, observed value ztI.e. the autoregressive model of the number of hysteretic variables at time t, lambdapIs a regression parameter between historical traffic flow data and influencing factors, p is the number of lagging variables at peak time of traffic flow, vtRepresenting a noise value of the white noise processing process of the historical traffic flow data at the time t;
and substituting each historical traffic flow data in the time sequence to be detected into the autoregressive model to obtain a corresponding observed value, and finishing white noise process processing.
5. The traffic flow prediction method based on algorithmic model analysis as defined in claim 1, wherein the traffic flow prediction method based on algorithmic model analysis further comprises calculating a traffic flow data variation trend, the calculating a traffic flow data variation trend comprising:
constructing an increase ratio index, wherein the expression is as follows:
in the formula, ziIndicates the predicted value of the traffic flow obtained by the current prediction, zi-1And the traffic flow data of the previous time period i-1 relative to the time period i where the current traffic flow predicted value is located.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011304652.6A CN112489422B (en) | 2020-11-19 | 2020-11-19 | Traffic flow prediction method based on algorithm model analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011304652.6A CN112489422B (en) | 2020-11-19 | 2020-11-19 | Traffic flow prediction method based on algorithm model analysis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112489422A true CN112489422A (en) | 2021-03-12 |
| CN112489422B CN112489422B (en) | 2022-11-04 |
Family
ID=74932116
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011304652.6A Active CN112489422B (en) | 2020-11-19 | 2020-11-19 | Traffic flow prediction method based on algorithm model analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112489422B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113345252A (en) * | 2021-06-08 | 2021-09-03 | 重庆大学 | Short-time prediction method and device for lower-channel flow of toll station |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105321345A (en) * | 2015-09-18 | 2016-02-10 | 浙江工业大学 | Road traffic flow prediction method based on ARIMA model and kalman filtering |
| CN108510741A (en) * | 2018-05-24 | 2018-09-07 | 浙江工业大学 | A kind of traffic flow forecasting method based on Conv1D-LSTM neural network structures |
| CN109377752A (en) * | 2018-10-19 | 2019-02-22 | 桂林电子科技大学 | Short-term traffic flow variation prediction method, apparatus, computer equipment and storage medium |
| CN110263860A (en) * | 2019-06-21 | 2019-09-20 | 广东工业大学 | A kind of freeway traffic flow prediction technique and device |
-
2020
- 2020-11-19 CN CN202011304652.6A patent/CN112489422B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105321345A (en) * | 2015-09-18 | 2016-02-10 | 浙江工业大学 | Road traffic flow prediction method based on ARIMA model and kalman filtering |
| CN108510741A (en) * | 2018-05-24 | 2018-09-07 | 浙江工业大学 | A kind of traffic flow forecasting method based on Conv1D-LSTM neural network structures |
| CN109377752A (en) * | 2018-10-19 | 2019-02-22 | 桂林电子科技大学 | Short-term traffic flow variation prediction method, apparatus, computer equipment and storage medium |
| CN110263860A (en) * | 2019-06-21 | 2019-09-20 | 广东工业大学 | A kind of freeway traffic flow prediction technique and device |
Non-Patent Citations (2)
| Title |
|---|
| 于明星: "基于时间序列ＡＲＩＭＡ模型的高职高专院校生师比预测", 《辽宁师专学报(自然科学版)》 * |
| 胡浩等: "基于组合预测方法的城市道路短时交通流预测", 《工业工程与管理》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113345252A (en) * | 2021-06-08 | 2021-09-03 | 重庆大学 | Short-time prediction method and device for lower-channel flow of toll station |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112489422B (en) | 2022-11-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111598148B (en) | Capacity evaluation method and device based on historical capacity similarity characteristics | |
| CN110517485B (en) | Short-term traffic flow prediction method based on time interval division | |
| CN113096388B (en) | Short-term traffic flow prediction method based on gradient lifting decision tree | |
| CN107563566B (en) | Inter-bus-station operation time interval prediction method based on support vector machine | |
| WO2009118987A1 (en) | Travel pattern information obtaining device, travel pattern information obtaining method, and travel pattern information obtaining program | |
| CN101706996A (en) | Method for identifying traffic status of express way based on information fusion | |
| CN110705774A (en) | Vehicle energy consumption analysis prediction method and system | |
| CN108734958B (en) | Traffic speed prediction method | |
| CN116013087A (en) | Traffic flow statistical method based on urban moving vehicle detection | |
| CN113470362A (en) | Urban road traffic jam space-time accurate discrimination method based on SVR-DEA model | |
| CN116524712A (en) | Highway congestion prediction method, system and device integrating space-time associated data | |
| CN112489422B (en) | Traffic flow prediction method based on algorithm model analysis | |
| AU2018215876B2 (en) | A method of predicting a traffic behaviour in a road system | |
| CN111626338B (en) | Cloud environment matching method, device, equipment and medium based on fusion classification model | |
| CN113449905A (en) | Traffic jam early warning method based on gated cyclic unit neural network | |
| CN116739376A (en) | Highway pavement preventive maintenance decision method based on data mining | |
| CN106875681B (en) | A kind of traffic congestion tendency prediction technique and system | |
| CN111862583B (en) | Traffic flow prediction method and device | |
| CN117094535B (en) | Artificial intelligence-based energy supply management method and system | |
| CN105761489A (en) | Cubic exponential smoothing optimal method of traffic flow prediction | |
| CN114463978B (en) | Data monitoring method based on track traffic information processing terminal | |
| CN115936242A (en) | Method and device for obtaining traceability relation data of air quality and traffic condition | |
| CN114973681A (en) | In-transit vehicle sensing method and device | |
| CN114281808A (en) | Traffic big data cleaning method, device, equipment and readable storage medium | |
| CN110120154B (en) | Traffic road condition prediction method using detector data under large-scale road network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |