CN110223510B - Multi-factor short-term traffic flow prediction method based on neural network LSTM - Google Patents
Multi-factor short-term traffic flow prediction method based on neural network LSTM Download PDFInfo
- Publication number
- CN110223510B CN110223510B CN201910334053.XA CN201910334053A CN110223510B CN 110223510 B CN110223510 B CN 110223510B CN 201910334053 A CN201910334053 A CN 201910334053A CN 110223510 B CN110223510 B CN 110223510B
- Authority
- CN
- China
- Prior art keywords
- traffic flow
- weather
- data
- holiday
- neural network
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 25
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 14
- 238000003062 neural network model Methods 0.000 claims abstract description 27
- 238000004140 cleaning Methods 0.000 claims abstract description 7
- 238000010606 normalization Methods 0.000 claims abstract description 4
- 238000012216 screening Methods 0.000 claims description 14
- 238000012549 training Methods 0.000 claims description 13
- 230000002159 abnormal effect Effects 0.000 claims description 12
- 238000007781 pre-processing Methods 0.000 claims description 8
- 238000011478 gradient descent method Methods 0.000 claims description 3
- 230000001902 propagating effect Effects 0.000 claims description 3
- 230000000694 effects Effects 0.000 description 21
- 238000011156 evaluation Methods 0.000 description 6
- 238000011160 research Methods 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 description 1
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 description 1
- 206010039203 Road traffic accident Diseases 0.000 description 1
- 238000011217 control strategy Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000006698 induction Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 239000004576 sand Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield 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
- G06N3/045—Combinations of networks
-
- 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
-
- 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
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Molecular Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Traffic Control Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
本发明属于交通工程领域，公开了一种基于神经网络LSTM的多因素短期车流量预测方法，步骤1：获取一段时间内的车流量数据，并对车流量数据进行预处理，得到短时交通流数据；步骤2：对短时交通流数据根据天气记录和节假日记录进行筛选，划分数据集；步骤3：进行数据清洗、数据重构和归一化；步骤4：建立LSTM神经网络模型，根据待预测日期的天气状况和节假日状况选择数据集，利用选择的数据集训练LSTM神经网络模型并调整LSTM参数，根据建立好的LSTM神经网络模型获得预测日期的车流量。本发明提出了更加细致化的思想，排除了其他因素，如天气因素、节假日因素等对交通流的影响，相对的提高了预测精度，使得对未来某时段的交通流预测更为准确有效。
The invention belongs to the field of traffic engineering, and discloses a multi-factor short-term traffic flow prediction method based on neural network LSTM. Data; Step 2: Screen the short-term traffic flow data according to weather records and holiday records, and divide the data set; Step 3: Carry out data cleaning, data reconstruction and normalization; Step 4: Establish an LSTM neural network model, according to the Select a data set for the weather conditions and holiday conditions of the predicted date, train the LSTM neural network model with the selected data set and adjust the LSTM parameters, and obtain the traffic flow on the predicted date according to the established LSTM neural network model. The invention proposes a more detailed idea, excludes the influence of other factors, such as weather factors, holiday factors, etc. on the traffic flow, relatively improves the prediction accuracy, and makes the traffic flow prediction for a certain period of time in the future more accurate and effective.
Description
Technical Field
The invention belongs to the field of traffic engineering, and particularly relates to a multi-factor short-term traffic flow prediction method based on a neural network LSTM.
Background
Traffic flow is used as a basis for upper-layer applications such as traffic management and intelligent traffic control, and the research of a short-time traffic flow prediction model based on time series is a research hotspot of scholars at home and abroad, so that a traffic control and induction system is a popular core course of ITS research. The key to realizing traffic guidance is to accurately predict the traffic flow, namely, the traffic flow condition of the next time period is predicted by establishing a proper model by utilizing the real-time and historical traffic flow data of the existing lane. Predictions of traffic flow may be divided into long-term predictions, medium-term predictions, and short-term predictions, which may serve different research fields. The short-time prediction can predict the traffic flow condition in a short time in the future, can meet the requirements of a traffic control system on real-time performance and accuracy, can obtain a correct control strategy, and has important practical significance for relieving traffic jam, reducing the exhaust emission of motor vehicles, reducing the incidence rate of traffic accidents and the like.
In the prediction of short-term traffic flows, two different types of data-driven prediction methods are frequently used: one is a method based on traditional statistical theory and the other is a method based on neural networks. The statistical-based methods mainly include kalman filter models, time series models, bayesian networks, and markov chains. The method is simple in theory and easy to understand, and the future prediction data and the past data have the same characteristics. However, most of the statistical-based methods are linear-based, and the traffic flow changes are non-linear and random, which deteriorates the performance of the statistical-based models. With the wide application of deep learning in the research fields of voice, image, natural language processing and the like, the neural network-based method has more and more powerful application value in the aspect of traffic big data mining.
Researchers have proposed short-time traffic flow prediction models based on deep belief network models, hybrid deep learning frameworks combining convolutional neural networks and long-short term memory networks, traffic prediction models combining convolutional neural networks and support vector regression classifiers, and the like.
An LSTM short-term traffic flow prediction model not only can more effectively deal with traffic flow nonlinearity and randomness, but also exhibits superior capability in interpreting time series predictions for a long period of time. LSTM can "remember" history information that is far from the current time point, and can automatically determine the optimal time interval of time-series data, thereby achieving good prediction effects. However, the traffic flow is also influenced by various factors such as weather conditions, holidays and the like, and when the LSTM model is used for prediction, the influence of various factors on the traffic flow is ignored, so that the effect of short-term prediction by the LSTM model is difficult to achieve the best effect.
Disclosure of Invention
The invention aims to provide a neural network LSTM model-based multi-factor traffic flow prediction method, which is used for solving the problem that the prediction effect of an LSTM prediction model in the prior art is poor.
In order to realize the task, the invention adopts the following technical scheme:
a multi-factor short-term traffic flow prediction method based on a neural network LSTM comprises the following steps:
step 1: acquiring traffic flow data within a period of time, and preprocessing the traffic flow data to obtain short-time traffic flow data;
step 2: screening the short-time traffic flow data obtained in the step 1 according to weather records and holiday records, taking short-time traffic flow data only in abnormal weather periods as a first data set, taking short-time traffic flow data only in holiday periods as a second data set, taking short-time traffic flow data in non-abnormal weather periods and holiday periods as a third data set, and taking short-time traffic flow data meeting both abnormal weather periods and holiday periods as a fourth data set;
and step 3: performing data cleaning, data reconstruction and normalization on the first data set, the second data set, the third data set and the fourth data set obtained in the step 2;
and 4, step 4: and (3) establishing an LSTM neural network model, selecting one of the normalized first data set, the normalized second data set, the normalized third data set and the normalized fourth data set obtained in the step (3) according to the weather condition and the holiday condition of the date to be predicted, training the LSTM neural network model by using the selected data set, adjusting LSTM parameters, and obtaining the traffic flow of the predicted date according to the established LSTM neural network model.
Further, step 1 comprises the following substeps:
and acquiring traffic flow data from a database of the highway toll station, preprocessing the traffic flow data, counting the number of vehicles every 15min to obtain short-time traffic flow data, wherein the traffic flow data comprises the starting time of the toll booth, the entrance network number, the entrance shift, the entrance vehicle type, the vehicle access date and time.
Further, the step 2 of screening the short-term traffic flow data obtained in the step 1 according to the weather record and the holiday record comprises the following steps:
(1) inquiring weather records, screening the short-term traffic flow data obtained in the step 1 according to the weather records, obtaining weather influence intensity factors of different weathers according to a formula I, dividing a time period in which the weather influence intensity factor is greater than a weather threshold value into abnormal weather time periods,
wherein, ViRepresenting traffic flow under different weather conditions, T representing total traffic flow in the month, i representing different weather, CiWeather influence coefficient, theta, representing different weatheriRepresenting the weather influence intensity factor, wherein the weather threshold is theta at sunny daysiTaking the value of (A);
(2) inquiring holiday records, screening the short-term traffic flow data obtained in the step 1 according to the holiday records, obtaining holiday influence coefficients according to a formula II, dividing the time period in which the holiday influence coefficients are larger than holiday threshold values into holiday time periods,
wherein E isjIndicates the traffic flow, beta, in the case of holidays of different festivalsjRepresenting holiday influence coefficients, j representing different holidays, the holiday threshold being beta when the holidays are not holidaysjThe value of (a).
Further, in step 3, a MInMaxScaler method is adopted for data cleaning and data reconstruction.
Further, training the LSTM neural network model and adjusting LSTM parameters using the selected data set in step 4, and obtaining the traffic flow of the predicted date according to the established LSTM neural network model specifically is as follows:
establishing a forgetting gate and an input gate in the LSTM neural network model, controlling the forgetting degree of the selected data set by using the forgetting gate, controlling the information of the predicted date by using the input gate, propagating a gradient value backwards by a gradient descent method, obtaining the optimal LSTM parameter after multiple iterations to further obtain the optimal LSTM neural network model, and obtaining the traffic flow of the predicted date by using the optimal LSTM neural network model.
Further, the traffic flow data acquired in the step 1 within a period of time is the traffic flow data within one month.
Compared with the prior art, the invention has the following technical characteristics:
1. the method is different from the conventional LSTM model for predicting the traffic flow, provides a more detailed thought, eliminates the influence of other factors such as weather factors, holiday factors and the like on the traffic flow, relatively improves the prediction precision, and enables the traffic flow prediction in a certain period of time in the future to be more accurate and effective.
2. The invention provides a data processing method for enabling data to be purer, and time periods corresponding to special weather and holidays are classified.
3. The invention classifies the data sets, selects more accurate data sets when predicting various conditions and improves the prediction precision.
Drawings
FIG. 1 is a flowchart of the process of the present invention;
FIG. 2(a) is the traffic flow line graph data in example 1
FIG. 2(b) is a graph showing the evaluation effect of the predicted date obtained after training in example 1;
FIG. 3(a) is the traffic flow line graph data in example 1 excluding the weather factor
FIG. 3(b) is a graph showing the evaluation effect of the predicted date obtained after training in example 1 excluding the weather factor;
FIG. 4(a) is the traffic flow line graph data in example 2
FIG. 4(b) is a graph showing the evaluation effect of the predicted date obtained after training in example 2;
the present invention will be described in further detail with reference to the accompanying drawings.
Detailed Description
The embodiment discloses a multi-factor prediction method based on a neural network LSTM, which specifically includes the following steps as shown in fig. 1:
step 1: acquiring traffic flow data within a period of time, and preprocessing the traffic flow data to obtain short-time traffic flow data;
step 2: screening the short-time traffic flow data obtained in the step 1 according to weather records and holiday records, taking short-time traffic flow data only in abnormal weather periods as a first data set, taking short-time traffic flow data only in holiday periods as a second data set, taking short-time traffic flow data in non-abnormal weather periods and holiday periods as a third data set, and taking short-time traffic flow data meeting both abnormal weather periods and holiday periods as a fourth data set;
and step 3: performing data cleaning, data reconstruction and normalization on the first data set, the second data set, the third data set and the fourth data set obtained in the step 2;
and 4, step 4: and (3) establishing an LSTM neural network model, selecting one of the normalized first data set, the normalized second data set, the normalized third data set and the normalized fourth data set obtained in the step (3) according to the weather condition and the holiday condition of the date to be predicted, training the LSTM neural network model by using the selected data set, adjusting LSTM parameters, and obtaining the traffic flow of the predicted date according to the established LSTM neural network model.
Specifically, step 1 includes the following substeps:
and acquiring traffic flow data from a database of the highway toll station, preprocessing the traffic flow data, counting the number of vehicles every 15min to obtain short-time traffic flow data, wherein the traffic flow data comprises the starting time of the toll booth, the entrance network number, the entrance shift, the entrance vehicle type, the vehicle access date and time.
Preferably, a select statement is adopted when the traffic flow data is acquired from the database of the highway toll station, the traffic flow data acquired after preprocessing is useful information, namely date and time, which represents the time when a vehicle enters the toll station, data1. sample (period) sum () is used for carrying out downsampling on the data every 15min to count the number of vehicles, and short-time traffic flow data every 15min is acquired through preprocessing at the moment.
Specifically, the step 2 of screening the short-term traffic flow data obtained in the step 1 according to the weather record and the holiday record comprises the following steps:
(1) inquiring weather records, screening the short-term traffic flow data obtained in the step 1 according to the weather records, obtaining weather influence intensity factors of different weathers according to a formula I, dividing a time period in which the weather influence intensity factor is greater than a weather threshold value into abnormal weather time periods,
wherein, ViIndicating traffic flow in different weather conditions, T indicating total traffic flow in the current month, i indicating different weather conditions, e.g. sunny, cloudy, light, heavy, medium, foggy, etc., CiWeather influence coefficient, theta, representing different weatheriRepresenting the weather influence intensity factor, wherein the weather threshold is theta at sunny daysiTaking the value of (A);
by analysis, for example, the traffic flow is obviously reduced in heavy rain weather, when CiThe value of (i represents heavy rainy day) is smaller than that of normal weather, so CiThe smaller the value of (A), the larger the influence of the corresponding weather condition on the traffic flow.
θiThe larger the weather condition is, the larger the influence of the weather condition on the traffic flow is. For example, taking the partial data at the exit of the toll station in month 6 in 2018 as an example, the total traffic flow T in month june is 855418, and the table is as follows:
by the above table, daily traffic volume V of each day is knowniWeather conditions and total traffic flow in the month are T855418, which is then determined by
Determining the influence coefficients Ci0.032606281, 0.032351435, 0.029995862 and 0.030846908 respectively, and considering the weather condition, the worse the weather (such as light rain to middle rain) is, the influence coefficient C is obtained by analysisiThe smaller. Then is further processed by
Separately determining the intensity factors thetaiThe values are 30.66893768, 30.91052993, 33.33793175 and 32.41816003 respectively, and the influence coefficient C can be obtained by analysisiThe smaller the intensity factor thetaiThe larger. Finally, the weather is divided into different influence levels according to the intensity factor. If the intensity factor is 30.91052993 in light rain, the intensity factor can be adjusted toIts impact is graded as 1; the intensity factor in rainy weather from light rain to medium rain is 33.33793175, and its effect can be rated as 2.
Therefore passing through thetaiThe value of (a) divides the influence of i weather (i can be taken in sunny days, cloudy days, rainy days, foggy days and the like) on the short-time traffic flow into 4 grades:
(1)0 represents no effect;
(2)1 represents a slight effect;
(3)2 represents an influence;
(4)3 represents a greater effect;
the method has the advantages that the purity of the traffic flow under the normal weather condition is guaranteed, the data under the weather which is only influenced are screened and stored in a classified mode.
| Weather conditions | Whether or not there is an influence | Whether to perform screening or not | Impact grade |
| In sunny days | Is free of | Whether or not | 0 |
| Cloudy day | Is free of | Whether or not | 0 |
| Small rain (cloudy turn little rain) | Is provided with | Is that | 1 |
| Rain (little rain turning rain) | Is provided with | Is that | 2 |
| Heavy rain (middle rain turning heavy rain) | Is provided with | Is that | 3 |
| Fog with large size | Is provided with | Is that | 3 |
| Haze | Is provided with | Is that | 3 |
If severe weather including medium snow, heavy snow, sand storm, rainstorm, typhoon and the like occurs, the data under the weather can be deleted, and the purity of the data is ensured.
(2) Inquiring holiday records, screening the short-term traffic flow data obtained in the step 1 according to the holiday records, obtaining holiday influence coefficients according to a formula II, dividing the time period in which the holiday influence coefficients are larger than holiday threshold values into holiday time periods,
wherein E isjIndicates the traffic flow, beta, in the case of holidays of different festivalsjRepresenting holiday influence coefficients, j representing different holidays, the holiday threshold being beta when the holidays are not holidaysjThe value of (a).
By analysis, for example, the traffic flow rate in the golden week of eleven is obviously increased in October, when Ej(at this moment, j represents eleven holidays) is larger than the value of the non-holiday, and the total traffic flow T in the month is constant, so EjThe larger the value of (A), the output betajThe larger the value, the larger the influence of the holiday on the traffic flow at this time. Number of holidays according to statutory holidays and season and beta of the holidayjThe impact level may be divided into 3 levels:
(1) level 1 represents a slight effect;
(2) level 2 represents influence;
(3) level 3 represents a greater impact;
in order to make the prediction of the traffic flow more accurate, if the data of the holidays is included in the data, the data is subjected to screening operation.
Further, in step 3, a MInMaxScaler method is adopted for data cleaning and data reconstruction.
Further, training the LSTM neural network model and adjusting LSTM parameters using the selected data set in step 4, and obtaining the traffic flow of the predicted date according to the established LSTM neural network model specifically is as follows:
establishing a forgetting gate and an input gate in the LSTM neural network model, controlling the forgetting degree of the selected data set by using the forgetting gate, controlling the information of the predicted date by using the input gate, propagating a gradient value backwards by a gradient descent method, obtaining the optimal LSTM parameter after multiple iterations to further obtain the optimal LSTM neural network model, and obtaining the traffic flow of the predicted date by using the optimal LSTM neural network model.
The invention also adopts a plurality of model evaluation indexes to compare to evaluate the quality of the prediction result, including the square percentage error (MAPE), the square absolute error (MAE), the Mean Square Error (MSE) and the R square.
Wherein, R square:
For the errors of square percentage error, square absolute error and mean square error, the smaller the error, the better the prediction effect. The R-squared is the ratio of the multiple regression sum of squares to the total sum of squares, which is a statistic that measures the degree of fit in the multiple regression equation, reflecting the ratio explained by the regression equation estimated in the variation of the dependent variable. The closer the square of R is to 1, the better the fit of the regression. In the invention, the square of R is mainly used as a main basis, and other evaluation indexes are used as references.
Example 1:
in the example, data of June in 2018 are collected once by 15Min, because the data required by the example are provided by a Xian toll station, the Xian belongs to a temperate zone semi-humid continental monsoon climate, the climate is clear in four seasons, the climate is mild, the rainfall is moderate, and most of the rainy season is concentrated in two seasons of spring and summer. Data were observed to be in the form of a small rain turning to a negative on three consecutive days, from month 6, month 16, to month 6, month 18. At this time, data from No. 6/month No. 1 to No. 18 are used as a training set, and No. 6/month No. 19 is used as a test set to predict the traffic flow condition of No. 6/month No. 19, and it is known from past weather data that No. 6/month No. 19 is also cloudy, the weather condition is different from the previous three days, and the test effect is shown in FIG. 2. And deleting the data from No. 6/month 16 to No. 6/month 18, excluding the condition that raindrops turn negative for three consecutive days, using the data after excluding the weather as a training set, and using the data from No. 6/month 19 as a test set to predict the traffic flow condition of No. 19, wherein the prediction effect is shown in fig. 3(a) and 3 (b). As can be seen by comparing fig. 2(a), 2(b) and 3(a), 3(b), the evaluation index mse, rmes, mae is significantly larger in fig. 2(b) than in fig. 3(b), and the R square in fig. 3(b) is about 1% larger than the R square in fig. 2(b), and the closer the R square is to 1, indicating the better degree of fit of regression. As can be seen from comparison of the prediction effects of fig. 2(a) and fig. 2(b) and fig. 3(a) and fig. 3(b), the weather factor may affect the prediction of the traffic flow, and the prediction effect may be better after the weather factor in the data is eliminated.
Example 2:
in this example, to take into account the effect of holidays on traffic flow predictions, the data in the 2 nd festival of 2018 were selected as the training set, since depending on the local climate conditions, there is less rain and snow in winter, which minimizes the effect of weather factors. Because the spring festival in month 2 is the most important day of the year, and the 15 days before and after the spring festival are the peak periods of spring return, the data of the spring festival in month 2 is selected to fully explain the influence of holidays on the traffic flow. The data of 2 months is used as a training set to be trained in the LSTM model, the data of 3 months No. 1 is used as a test set to be predicted, and the result is shown in fig. 4(a) and 4(b), wherein mse and rmse are larger, and R square is smaller than 0.92, the neural network LSTM model shows that the model has better prediction effect, and the R square is generally larger than 0.93. The reason for this is that since the entire 2 months is the most busy period of spring transportation and the 3 months of spring transportation has ended, the data of holiday 2 months is used to predict the data of 3 months 1, and the prediction effect becomes poor. The influence of holiday factors on the predicted traffic flow of the LSTM model is illustrated by the example.
Claims (5)
1. A multi-factor short-term traffic flow prediction method based on a neural network LSTM is characterized by comprising the following steps:
step 1: acquiring traffic flow data within a period of time, and preprocessing the traffic flow data to obtain short-time traffic flow data;
step 2: screening the short-time traffic flow data obtained in the step 1 according to weather records and holiday records, taking short-time traffic flow data only in abnormal weather periods as a first data set, taking short-time traffic flow data only in holiday periods as a second data set, taking short-time traffic flow data in non-abnormal weather periods and holiday periods as a third data set, and taking short-time traffic flow data meeting both abnormal weather periods and holiday periods as a fourth data set;
the step 2 of screening the short-term traffic flow data obtained in the step 1 according to the weather record and the holiday record comprises the following steps:
(1) inquiring weather records, screening the short-term traffic flow data obtained in the step 1 according to the weather records, obtaining weather influence intensity factors of different weathers according to a formula I, dividing a time period in which the weather influence intensity factor is greater than a weather threshold value into abnormal weather time periods,
wherein, ViRepresenting traffic flow under different weather conditions, T representing total traffic flow in the month, i representing different weather, CiWeather influence coefficient, theta, representing different weatheriRepresenting the weather influence intensity factor, wherein the weather threshold is theta at sunny daysiTaking the value of (A);
(2) inquiring holiday records, screening the short-term traffic flow data obtained in the step 1 according to the holiday records, obtaining holiday influence coefficients according to a formula II, dividing the time period in which the holiday influence coefficients are larger than holiday threshold values into holiday time periods,
wherein E isjIndicates the traffic flow, beta, in the case of holidays of different festivalsjRepresenting holiday influence coefficients, j representing different holidays, the holiday threshold being beta when the holidays are not holidaysjTaking the value of (A);
and step 3: performing data cleaning, data reconstruction and normalization on the first data set, the second data set, the third data set and the fourth data set obtained in the step 2;
and 4, step 4: and (3) establishing an LSTM neural network model, selecting one of the normalized first data set, the normalized second data set, the normalized third data set and the normalized fourth data set obtained in the step (3) according to the weather condition and the holiday condition of the date to be predicted, training the LSTM neural network model by using the selected data set, adjusting LSTM parameters, and obtaining the traffic flow of the predicted date according to the established LSTM neural network model.
2. The neural network LSTM based multi-factor short term traffic prediction method of claim 1, wherein step 1 comprises the sub-steps of:
and acquiring traffic flow data from a database of the highway toll station, preprocessing the traffic flow data, counting the number of vehicles every 15min to obtain short-time traffic flow data, wherein the traffic flow data comprises the starting time of the toll booth, the entrance network number, the entrance shift, the entrance vehicle type, the vehicle access date and time.
3. The method of claim 1, wherein the MINMAxScaler method is used in step 3 for data cleaning and data reconstruction.
4. The method for predicting short-term traffic flow based on LSTM neural network of claim 1, wherein the training of LSTM neural network model and the adjustment of LSTM parameters using the selected data set in step 4, the traffic flow for obtaining the predicted date according to the established LSTM neural network model is specifically:
establishing a forgetting gate and an input gate in the LSTM neural network model, controlling the forgetting degree of the selected data set by using the forgetting gate, controlling the information of the predicted date by using the input gate, propagating a gradient value backwards by a gradient descent method, obtaining the optimal LSTM parameter after multiple iterations to further obtain the optimal LSTM neural network model, and obtaining the traffic flow of the predicted date by using the optimal LSTM neural network model.
5. The neural network LSTM-based multi-factor short-term traffic flow prediction method as claimed in claim 1, wherein the traffic flow data obtained in step 1 is the traffic flow data in a month.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910334053.XA CN110223510B (en) | 2019-04-24 | 2019-04-24 | Multi-factor short-term traffic flow prediction method based on neural network LSTM |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910334053.XA CN110223510B (en) | 2019-04-24 | 2019-04-24 | Multi-factor short-term traffic flow prediction method based on neural network LSTM |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110223510A CN110223510A (en) | 2019-09-10 |
| CN110223510B true CN110223510B (en) | 2021-03-26 |
Family
ID=67819971
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910334053.XA Active CN110223510B (en) | 2019-04-24 | 2019-04-24 | Multi-factor short-term traffic flow prediction method based on neural network LSTM |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110223510B (en) |
Families Citing this family (22)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110837888A (en) * | 2019-11-13 | 2020-02-25 | 大连理工大学 | Traffic missing data completion method based on bidirectional cyclic neural network |
| CN110847272B (en) * | 2019-11-21 | 2021-08-03 | 中交疏浚技术装备国家工程研究中心有限公司 | Intelligent dredging control method for cutter suction dredger |
| CN111027461B (en) * | 2019-12-06 | 2022-04-29 | 长安大学 | Vehicle track prediction method based on multi-dimensional single-step LSTM network |
| CN111144648B (en) * | 2019-12-25 | 2023-11-24 | 中国联合网络通信集团有限公司 | People flow prediction equipment and method |
| CN111239014B (en) * | 2020-01-09 | 2022-10-21 | 河北工业大学 | A detection method, device, equipment and medium for PM2.5 in air |
| CN111260122A (en) * | 2020-01-13 | 2020-06-09 | 重庆首讯科技股份有限公司 | Method and device for predicting traffic flow on expressway |
| CN111445694B (en) * | 2020-03-04 | 2022-02-01 | 青岛海信网络科技股份有限公司 | Festival and holiday traffic scheduling method and device based on traffic flow prediction |
| CN111524348A (en) * | 2020-04-14 | 2020-08-11 | 长安大学 | A long-term and short-term traffic flow prediction model and method |
| CN111768622A (en) * | 2020-06-23 | 2020-10-13 | 南通大学 | A Short-term Traffic Prediction Method Based on Improved Grey Wolf Algorithm |
| CN111882878B (en) * | 2020-09-02 | 2021-07-02 | 烟台大学 | A method for maximizing critical road capacity based on traffic flow prediction |
| CN112116140B (en) * | 2020-09-10 | 2022-09-20 | 同济大学 | Building energy consumption prediction method based on twin model |
| CN112382087B (en) * | 2020-11-05 | 2021-10-08 | 杭州易上科技有限公司 | Traffic jam prediction method |
| CN112949893A (en) * | 2020-11-18 | 2021-06-11 | 安徽师范大学 | Soybean reserve early warning method based on improved RNN |
| CN112820103B (en) * | 2020-12-31 | 2022-09-06 | 山东奥邦交通设施工程有限公司 | Integrated dynamic and static combined marker management and control method and system |
| CN112862177B (en) * | 2021-02-02 | 2024-01-19 | 湖南大学 | Urban area aggregation degree prediction method, device and medium based on deep neural network |
| CN113674524A (en) * | 2021-07-28 | 2021-11-19 | 长安大学 | Multi-scale short-term traffic flow prediction modeling, prediction method and system based on LSTM-GASVR |
| CN113971479B (en) * | 2021-10-01 | 2024-11-26 | 浙江省机电设计研究院有限公司 | A method for predicting daily traffic flow parameters integrating the impact of holidays |
| CN114255591A (en) * | 2021-12-17 | 2022-03-29 | 重庆中信科信息技术有限公司 | Short-term traffic flow prediction method and device considering space-time correlation and storage medium |
| CN115063975B (en) * | 2022-06-09 | 2023-07-11 | 长安大学 | Short-term traffic flow data prediction method, system, computer equipment and storage medium |
| CN115909737A (en) * | 2022-11-18 | 2023-04-04 | 河北万方中天科技有限公司 | Traffic timing method and device, traffic control platform, readable storage medium |
| CN115762161A (en) * | 2022-11-18 | 2023-03-07 | 河北万方中天科技有限公司 | Traffic flow control method and device, traffic control platform, readable storage medium |
| CN118591061B (en) * | 2024-06-12 | 2025-02-28 | 杭州立方控股股份有限公司 | Energy-saving access control system |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101620781A (en) * | 2008-06-30 | 2010-01-06 | 株式会社查纳位资讯情报 | System for predicting passenger information and system for searching passenger information and method thereof |
| CN103956043A (en) * | 2014-04-29 | 2014-07-30 | 南京理工大学 | Auxiliary vehicle traveling path system based on mobile terminal |
| JP2017220000A (en) * | 2016-06-07 | 2017-12-14 | 日本電信電話株式会社 | Flow rate prediction device, pattern estimation device, flow rate prediction method, pattern estimation method and program |
| CN107610469A (en) * | 2017-10-13 | 2018-01-19 | 北京工业大学 | A kind of day dimension regional traffic index forecasting method for considering multifactor impact |
| CN108417034A (en) * | 2018-03-23 | 2018-08-17 | 四川高路交通信息工程有限公司 | A kind of highway festivals or holidays vehicle flowrate prediction technique based on Multi-variable Grey Model |
| CN109658695A (en) * | 2019-01-02 | 2019-04-19 | 华南理工大学 | A kind of multifactor Short-time Traffic Flow Forecasting Methods |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP4175312B2 (en) * | 2004-09-17 | 2008-11-05 | 株式会社日立製作所 | Traffic information prediction device |
-
2019
- 2019-04-24 CN CN201910334053.XA patent/CN110223510B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101620781A (en) * | 2008-06-30 | 2010-01-06 | 株式会社查纳位资讯情报 | System for predicting passenger information and system for searching passenger information and method thereof |
| CN103956043A (en) * | 2014-04-29 | 2014-07-30 | 南京理工大学 | Auxiliary vehicle traveling path system based on mobile terminal |
| JP2017220000A (en) * | 2016-06-07 | 2017-12-14 | 日本電信電話株式会社 | Flow rate prediction device, pattern estimation device, flow rate prediction method, pattern estimation method and program |
| CN107610469A (en) * | 2017-10-13 | 2018-01-19 | 北京工业大学 | A kind of day dimension regional traffic index forecasting method for considering multifactor impact |
| CN108417034A (en) * | 2018-03-23 | 2018-08-17 | 四川高路交通信息工程有限公司 | A kind of highway festivals or holidays vehicle flowrate prediction technique based on Multi-variable Grey Model |
| CN109658695A (en) * | 2019-01-02 | 2019-04-19 | 华南理工大学 | A kind of multifactor Short-time Traffic Flow Forecasting Methods |
Non-Patent Citations (2)
| Title |
|---|
| Long Short-Term Memory Recurrent Neural Network for Urban Traffic Prediction: A Case Study of Seoul;Yong-Ju Lee等;《IEEE》;20181231;全文 * |
| 基于深度学习的交通流量预测研究;邓烜堃等;《计算机工程与应用》;20180516;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110223510A (en) | 2019-09-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110223510B (en) | Multi-factor short-term traffic flow prediction method based on neural network LSTM | |
| CN110503833B (en) | An on-ramp linkage control method based on deep residual network model | |
| CN103984993B (en) | Rail transit passenger flow OD distribution real-time speculation method | |
| CN106951976B (en) | Bus passenger flow prediction method based on mode classification | |
| CN112289034A (en) | Deep neural network robust traffic prediction method based on multi-mode space-time data | |
| CN110956807B (en) | Highway flow prediction method based on combination of multi-source data and sliding window | |
| CN108417032B (en) | An analysis and forecasting method of on-street parking demand in urban central area | |
| CN114202103A (en) | Machine learning-based holiday scenic spot passenger flow prediction method | |
| CN111462485A (en) | Traffic intersection congestion prediction method based on machine learning | |
| CN111931978A (en) | Urban rail transit passenger flow state prediction method based on space-time characteristics | |
| Zhang et al. | Analysis of rainfall effects on road travel speed in Beijing, China | |
| CN109035762B (en) | A Traffic Speed Prediction Method Based on Spatio-temporal Width Learning | |
| CN110889558A (en) | Road condition prediction method and device | |
| CN107180270A (en) | Passenger flow forecasting and system | |
| CN116384635B (en) | Green Ecological City Information Management System Based on Big Data | |
| CN106845714A (en) | A kind of monthly passenger flow method of ARIMA model prediction urban track traffics based on seasonal index number | |
| CN117809454A (en) | Expressway running state studying and judging method integrating intercommunication and main line data | |
| WO2021115320A1 (en) | Traffic evaluation method and system | |
| Makarova et al. | Improving Road Safety by Affecting Negative Factors. | |
| Wang et al. | The effects of weather on passenger flow of urban rail transit | |
| CN114723103A (en) | Highway toll income prediction method, system, equipment and storage medium | |
| CN113537596A (en) | Short-time passenger flow prediction method for new line station of urban rail transit | |
| CN106503848A (en) | The load forecasting method of many small power station's bulk sale area power grids | |
| Liu et al. | A MRT daily passenger flow prediction model with different combinations of influential factors | |
| CN114091772B (en) | Multi-angle fusion road traffic flow prediction method based on encoder and decoder |
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 |