Prediction method for urban road traffic congestion index

Abstract

The invention discloses a method for predicting an urban road traffic congestion index, and the method comprises the following steps: obtaining congestion index data of a target road section, and decomposing the congestion index data to obtain intrinsic mode functions and residual terms of different frequencies; training each intrinsic mode function component to obtain prediction results of intrinsic mode functions of different frequencies; superposing and constructing prediction results of all the subsequences, and integrating prediction values of the congestion indexes; the predicted value is verified based on the evaluation index, and the degree of deviation between the predicted value and the true value is quantified. According to the prediction method, the high-quality congestion index data is obtained, the time sequence information of the sequence data is fully utilized, and the prediction accuracy is improved; the prediction results of all the subsequences are superposed and reconstructed, the consistency and continuity of the prediction results are kept, finally, the effects of mining complex fluctuation of data, accurately predicting the congestion index at the peak moment and improving the prediction precision of the congestion index are achieved, and the method brings brand new development for relieving the urban traffic congestion problem and improves the prediction accuracy.

Description

Technical Field

[0001] The present application relates to the field of intelligent transportation, and specifically to a method for predicting urban road traffic congestion index. Background Art

[0002] In recent years, with the continuous advancement of urban modernization, traffic congestion has become a bottleneck problem in the development of many cities. How to effectively analyze historical congestion index data, predict the congestion index, understand the causes and bottleneck areas of urban traffic congestion, and provide data support for the formulation of targeted traffic management measures are important issues that need to be solved at present.

[0003] In the existing research on traffic congestion index prediction, four methods are usually used to predict the road congestion index: statistical analysis prediction model, nonlinear theoretical model, machine learning prediction model and combined prediction model. Although existing research has made many improvements in the prediction of congestion index, there are still areas to be improved. First, existing research mainly uses a single prediction model or multiple machine learning combination models to predict the road congestion index, which has certain shortcomings in mining the complex fluctuation structure of the congestion index; secondly, the prediction accuracy of traditional methods such as nonlinear regression needs to be improved; in addition, the existing prediction model has a good prediction effect when the congestion index series is stable, but the prediction accuracy needs to be improved when the peak congestion index fluctuates significantly. Summary of the invention

[0004] In response to the above-mentioned problems existing in the prior art, the present application provides a method for predicting the urban road traffic congestion index, which mines the complex fluctuations of data and accurately predicts the congestion index during peak hours, thereby improving the prediction accuracy of the congestion index and providing a new prediction method for alleviating the problem of urban traffic congestion.

[0005] To achieve the above purpose, the present application adopts a technical solution: a method for predicting urban road traffic congestion index, comprising the following steps:

[0006] Obtain the congestion index data of the target road section and decompose it to obtain the intrinsic mode functions and residual terms of different frequencies;

[0007] Train each intrinsic mode function component to obtain the prediction results of the intrinsic mode functions of different frequencies;

[0008] The prediction results of each subsequence are superimposed and constructed to integrate the predicted value of the congestion index;

[0009] The predicted value is verified based on the evaluation index, and the deviation between the predicted value and the true value is quantified.

[0010] Among them, obtaining the congestion index data of the target road section includes:

[0011] Obtain the historical data of the congestion index of the target road section and pre-process the historical data of the congestion index;

[0012] Descriptive statistical analysis is performed on the processed data to obtain congestion index data.

[0013] Among them, the preprocessing includes the processing of outliers, duplicate values ​​and missing values ​​in the historical data of the congestion index, and the descriptive statistics include the analysis of mean, variance, skewness, kurtosis, as well as normality test and serial autocorrelation test.

[0014] Among them, the inherent mode functions and residual terms of different frequencies decomposed include:

[0015] Adding N times the adaptive noise of the same length to the original signal sequence to obtain a new signal sequence, decomposing the new sequence N times based on empirical mode decomposition, and the average value is the first characteristic mode function;

[0016] Obtain a first margin sequence, and calculate a second characteristic mode function based on the margin sequence;

[0017] Repeat the calculation to obtain the marginal sequences of the k+1 stage and the kth stage and the k+1 characteristic mode components;

[0018] The calculation is repeated until the extreme value point of the margin sequence is no more than 2, and K intrinsic mode functions and residual terms are obtained.

[0019] The new signal sequence x i (t), first characteristic mode function IMF 1 The expression is:

[0020] x i (t) = x(t) + εω i (t);

[0021]

[0022] Where x(t) is the original signal sequence, ε is the adaptive coefficient, ω i (t) is the noise sequence.

[0023] Among them, the first margin sequence r 1 (t), the second characteristic mode function IMF 2 The expression is:

[0024] r 1 (t) = x(t) - IMF 1 (t);

[0025]

[0026] Among them, the marginal sequence r of the k+1 stage and the kth stage k(t) and k+1 eigenmode components IMF k+1 ,

[0027] The expression is:

[0028] r k (t) = r k-1 (t)-IMF k (t);

[0029]

[0030] Among them, the expressions of K intrinsic mode functions and residual terms are:

[0031]

[0032] Among them, training each intrinsic mode function component includes:

[0033] By resetting the gate, the weight matrix W of Xr Get the output R of the reset gate t , the expression is:

[0034] R t =σ(X t W xr +H t-1 W hr +b r );

[0035] Where R t represents the output of the reset gate at time t; X t Represents the current input; H t-1 represents the hidden state of the previous moment; δ represents the sigmoid function; W xr , W hr represents the weight parameter; b r represents the deviation parameter;

[0036] By updating the gate, the hidden state H of the previous layer is t-1 Part of the data information is passed to the next layer for training, and the expression is:

[0037] Z t =σ(X t W xz +H t-1 W hz +b z

[0038] Where Z t represents the output of the update gate at time t; δ represents the sigmoid function; W xz , W hz represents the weight parameter; b z represents the deviation parameter;

[0039] Use the output of the reset gate at time t to multiply the hidden state at time t-1 element by element, combine the calculated value with the input at time t, and get the candidate hidden state The expression is:

[0040]

[0041] in represents the candidate hidden state at time t; W xh , W hh represents the weight parameter; b h represents the bias parameter; ☉ represents element-wise multiplication;

[0042] Based on the update gate Z t For H t-1 and Make a combination to get the hidden state H at time t t , the expression is:

[0043]

[0044] Among them, H t-1 Represents the hidden state of the previous moment; H t Represents the current hidden state; Represents the current candidate hidden state; Z t Represents the output of the current update gate; ☉ represents element-wise multiplication;

[0045] Different weights are assigned according to the correlation of different positions in the input sequence. The expression is:

[0046]

[0047] Where Q represents the query vector; K represents the key vector; V represents the value vector;

[0048] Among them, the evaluation indicators include determination coefficient, mean absolute percentage error and mean absolute error.

[0049] Beneficial effects of this application:

[0050] The present application provides a method for predicting urban road traffic congestion index, which obtains high-quality congestion index data, accurately and comprehensively captures the frequency characteristics of the congestion index data, and then makes full use of the time series information of the sequence data to improve the prediction accuracy; the prediction results of each subsequence are superimposed and reconstructed to maintain the consistency and continuity of the prediction results, and finally achieves the effect of mining the complex fluctuations of the data while accurately predicting the congestion index at peak times, thereby improving the prediction accuracy of the congestion index. The method brings new development to alleviate the problem of urban traffic congestion and improves the prediction accuracy, and has positive application value to urban traffic management and planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 A schematic diagram of a flow chart of a prediction method according to an embodiment of the present application;

[0052] Figure 2 A schematic diagram of a prediction model of a prediction method according to an embodiment of the present application;

[0053] Figure 3 This is a GRU network structure diagram of the prediction method shown in an embodiment of the present application.

[0054] Figure 4 This is a GRU prediction flowchart of the prediction method shown in an embodiment of the present application. DETAILED DESCRIPTION

[0055] The embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to make the present application more thorough and complete, and to fully convey the scope of the present application to those skilled in the art.

[0056] It should be understood that although the terms "first", "second", "third", etc. may be used in this application to describe various information, this information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of this application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0057] In the description of the present application, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present application.

[0058] Unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0059] Embodiment 1

[0060] A prediction method for urban road traffic congestion index based on the gated recurrent neural network attention model of CEEMDAN-GRU-ATTENTION complex empirical mode decomposition. CEEMDAN-GRU-ATTENTION is a deep learning model that combines CEEMDAN complete set empirical mode decomposition with adaptive noise, GRU gated recurrent unit and Attention mechanism. CEEMDAN is used to decompose and denoise time series data, and then the GRU network is used to model long-term dependencies, and finally the Attention mechanism is used to dynamically focus on important feature information. By combining these three components, the CEEMDAN-GRU-ATTENTION model efficiently extracts complex features in time series data for more accurate prediction and analysis.

[0061] like Figure 1-4 A method for predicting urban road traffic congestion index shown in the figure comprises the following steps:

[0062] S1. Obtain the congestion index data of the target road section and decompose it to obtain the intrinsic mode functions and residual terms of different frequencies. The following steps are included:

[0063] S11. Data preprocessing and analysis: Obtain the historical data of the congestion index of the road section, and preprocess the congestion index data. Data preprocessing includes processing outliers, duplicate values, and missing values, removing outliers and duplicate values ​​in the data, and interpolating the missing parts. Perform descriptive statistical analysis on the processed data; the descriptive statistics of the data include analysis of the mean, variance, skewness, and kurtosis, and perform Jarque Bera normality test and Ljung-Box test to determine whether the data set is close to normal distribution and whether the autocorrelation in the time series data is significant.

[0064] S12. Data decomposition: The CEEMDAN model is used to decompose the congestion index data to obtain the intrinsic mode function (IMF) and residual terms of different frequencies.

[0065] The specific method is:

[0066] S121, obtain a new signal sequence x by adding N times the adaptive noise of the same length to the original signal sequence x(t) i (t). Use EMD to decompose the new sequence N times, and the average value is the first characteristic mode function IMF 1 , the expression is as follows:

[0067] x i (t) = x(t) + εω i (t

[0068]

[0069] Where ε is the adaptive coefficient and is the noise sequence.

[0070] S122. Calculate the first margin sequence r 1 (t), the expression is as follows:

[0071] r 1 (t) = x(t) - IMF 1 (t)

[0072] S123, based on S202, calculate the second characteristic mode function IMF 2 , the expression is as follows:

[0073]

[0074] S124. Repeat the calculation to obtain the marginal sequence r of the k+1 stage and the kth stage k (t) and k+1 eigenmode components IMF k+1 , the expression is as follows:

[0075] r k (t) = r k-1 (t)-IMF k (t

[0076]

[0077] S125, repeat the above steps until the extreme value of the margin sequence is no more than 2. Through the above steps, the original sequence can be decomposed into K characteristic mode functions and a margin sequence, and the expression is as follows:

[0078]

[0079] In the present invention, the original nonlinear and non-stationary time series data is decomposed into 15 groups of IMF components and a residual term R (n) .

[0080] S2. Train each intrinsic mode function component to obtain prediction results of intrinsic mode functions of different frequencies.

[0081] A congestion index prediction model based on GRU-ATTENTION is constructed, and each IMF intrinsic modal function component is trained and predicted to obtain the prediction results of IMFs of different frequencies. Among them, the activation function is selected as ReLU, which stands for rectified linear unit, and the expression is as follows:

[0082] f(x)=max(0,x)

[0083] The loss function is the mean square error (MSE), which is expressed as follows:

[0084]

[0085] Where n represents the total number of samples; y represents the true value; Represents the predicted value.

[0086] The Adam optimizer is selected as the optimizer, and the learning rate is set to 0.0003 to avoid overfitting of the model.

[0087] The operation steps of the GRU model are as follows:

[0088] Step 1: Clean the initial road congestion index data, including deleting duplicate values, identifying and repairing outliers and missing values, and normalizing the data.

[0089] Step 2: Initialize the network and set the basic initial parameters of the GRU model, including the number of hidden layer units, input step size, number of iterations, and forgetting rate;

[0090] Step 3: Divide the road congestion index dataset into a training set and a validation set in a ratio of 8:2;

[0091] Step 4: Perform iterative training of the GRU model. The data set will output the predicted value in the GRU model, and calculate the error between the model predicted value and the true value;

[0092] Step 5: Calculate the loss function and use the Adam optimizer to continuously improve the model parameters to make the predicted value output by the model more accurate

[0093] Step 6: Based on the saved loss history, draw the loss graphs of the training set and the validation set.

[0094] Step 7: Output the evaluation index of the model, and analyze the prediction effect and accuracy of the GRU model based on the numerical value of the evaluation index.

[0095] Step 8: Draw a comparison curve between the predicted data and the actual data.

[0096] Based on this, the specific method of S3 is:

[0097] S21, by resetting the gate, the weight matrix W of Xr Perform element-level multiplication, and then process it through the sigmoid function to get the output R of the reset gate. t , the expression is as follows:

[0098] R t =σ(X t W xr +H t-1 W hr +b r

[0099] Where R t Represents the output of the reset gate at any time; X t Represents the current input; H t-1 represents the hidden state of the previous moment; δ represents the sigmoid function; W xr , W hr represents the weight parameter; b r represents the deviation parameter.

[0100] S22, through the update gate, the hidden state H of the previous layer t-1 Part of the data information is passed to the next layer for training. The expression is as follows:

[0101] Z t =σ(X t W xz +H t-1 W hz +b z

[0102] Where Z t represents the output of the update gate at time t; δ represents the sigmoid function; W xz , W hz represents the weight parameter; b z represents the deviation parameter.

[0103] S23. Use the output of the reset gate at time t to perform element-by-element multiplication with the hidden state at time t-1. If the calculated value is approximately 0, it indicates that the hidden state information at time t-1 needs to be selectively ignored or "forgotten". On the contrary, when the calculated value is approximately 1, it means that the hidden state information at time t-1 should be stored. After performing unit-by-unit multiplication, the calculated value is merged with the input at time t. Finally, a suitable activation function is set in the fully connected layer to process the result obtained in the previous step to obtain the candidate hidden state. The expression is as follows:

[0104]

[0105] in represents the candidate hidden state at time t; W xh , W hh represents the weight parameter; b h represents the bias parameter; ☉ represents element-wise multiplication.

[0106] S24, use update gate Z t For H t-1 and Make a combination to get the hidden state H at time t t , the expression is as follows:

[0107]

[0108] Among them, H t-1 Represents the hidden state of the previous moment; H t Represents the current hidden state; Represents the current candidate hidden state; Z t Represents the output of the current update gate; ☉ represents element-wise multiplication.

[0109] S25. Combined with the self-attention mechanism, the model can assign different weights according to the relevance of different positions in the input sequence. The expression is as follows:

[0110]

[0111] Where Q represents the query vector; K represents the key vector; and V represents the value vector.

[0112] S3. The prediction results of each subsequence are superimposed and constructed to integrate the prediction value of the congestion index.

[0113] According to the intrinsic mode functions (IMFs) and residual terms of different frequencies obtained by CEEMDAN decomposition, each IMF and residual term is predicted to obtain the prediction results of IMFs of different frequencies. The prediction results of each IMF are added to the residual term to obtain the superposition results of each IMF and residual term. The superposition results of each IMF and residual term are added to obtain the predicted value of the congestion index.

[0114] Through this superposition and reconstruction process, the predicted IMF and residual terms of different frequencies are combined to obtain the overall congestion index prediction result, which comprehensively considers the influence of different frequency components and improves the prediction precision and accuracy of the congestion index.

[0115] When performing overlay reconstruction, the time sequence of each IMF and residual item should be kept consistent to ensure the continuity and accuracy of the prediction results. Therefore, the correct overlay order and the weight distribution of each IMF and residual item are very important. Appropriate algorithms or rules can be used to determine the appropriate overlay weights to obtain the best prediction results.

[0116] S4. Verify the predicted value based on the evaluation index and quantify the deviation between the predicted value and the true value.

[0117] The selected evaluation indicators include the coefficient of determination (R 2 ), mean absolute percentage error (MAPE), and mean absolute error (MAE).

[0118] Coefficient of determination R 2 , also known as the degree of fit, reflects the accuracy of the model fitting the data, R 2 The value range of is generally [0,1]. The closer its value is to 1, the better the model fitting effect is. The expression is as follows:

[0119]

[0120] The mean absolute error MAE is the average value of the absolute error, which can well reflect the actual situation of the prediction value error. The expression is as follows:

[0121]

[0122] The mean absolute percentage error MAPE represents the accuracy of the model. The smaller its value is, the better the accuracy of the prediction model is. The expression is as follows:

[0123]

[0124] Where n represents the total number of samples; y i represents the i-th true value; represents the i-th predicted value; Represents the average of i true values.

[0125] Embodiment 2

[0126] The road congestion index data of the target road section in the past three months were collected. The total amount of data was about 13,000, and the time granularity of each data was 10 minutes. The data set included six indicators such as congestion index, commuting time, and free flow speed.

[0127] In the data preprocessing stage, outliers and duplicate values ​​in the original data were deleted, and then missing values ​​were interpolated. The processed data were analyzed, including descriptive statistical analysis, Jarque Bera normality test, and Ljung-Box test.

[0128] In the data decomposition stage, the preprocessed data is decomposed by the CEEMDAN model to obtain 15 IMF components and one residual term, and descriptive statistics are performed one by one. Different IMF components represent the characteristics of the original signal at different time scales, and the residual reflects the trend of the original congestion index series.

[0129] In the model training and prediction stage, a GRU-ATTENTION model is constructed to mine the time scale features contained in each subsequence, and each sequence is trained and predicted separately to obtain the prediction results of the neural network model for different subsequences.

[0130] In the superposition and reconstruction stage, the prediction results of each subsequence are superimposed and reconstructed, and the prediction value of the congestion index sequence of the integrated prediction results is obtained.

[0131] Evaluation and analysis phase, through the output of R 2 , MAE and MAPE are used to analyze the performance of the constructed network model. The fitting degree of the final model is R 2 It can reach 94.4%, and the mean absolute percentage error MAPE can reach 6.3%, that is, the accuracy of the model can reach 93.7%, and the mean absolute error MAE of the model is only 0.11. Therefore, it can be seen from the output indicators that the constructed CEEMDAN-GRU-ATTENTION model has a good prediction effect on the road congestion index.

[0132] Finally, it should be noted that, in this article, relationships such as first and second, etc. are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms include, include or any other variations are intended to cover non-exclusive inclusion, so that a process, method, article or device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0133] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0134] The embodiments of the present application have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A method for predicting urban road traffic congestion index, characterized in that: The following steps are involved: Obtain the congestion index data of the target road section and decompose it to obtain the intrinsic mode functions and residual terms of different frequencies; Training each intrinsic mode function component to obtain prediction results of the intrinsic mode function at different frequencies; The prediction results of each subsequence are superimposed to construct a prediction value of the integrated congestion index; The predicted value is verified based on the evaluation index, and the deviation between the predicted value and the true value is quantified.

2. The method for predicting urban road traffic congestion index according to claim 1, characterized in that: The obtaining of the congestion index data of the target road section includes: Acquire historical congestion index data of the target road section, and preprocess the historical congestion index data; Descriptive statistical analysis is performed on the processed data to obtain the congestion index data.

3. The method for predicting urban road traffic congestion index according to claim 2, characterized in that: The preprocessing includes processing abnormal values, duplicate values ​​and missing values ​​of the congestion index historical data, and the descriptive statistics include analysis of mean, variance, skewness, kurtosis, normality test and sequence autocorrelation test.

4. The method for predicting urban road traffic congestion index according to claim 1, characterized in that: The decomposition results in the following natural mode functions and residual terms of different frequencies: Adding N times the adaptive noise of the same length to the original signal sequence to obtain a new signal sequence, decomposing the new sequence N times based on empirical mode decomposition, and the average value is the first characteristic mode function; Obtaining a first margin sequence, and calculating a second characteristic mode function based on the margin sequence; Repeat the calculation to obtain the marginal sequences of the k+1 stage and the kth stage and the k+1 characteristic mode components; The calculation is repeated until the extreme value point of the margin sequence is no more than 2, thereby obtaining K intrinsic mode functions and the residual terms.

5. The method for predicting urban road traffic congestion index according to claim 4, characterized in that: The new signal sequence x i (t), the expression of the first characteristic mode function IMF1 is: x i (t)=x(t)+εω i (t); Where x(t) is the original signal sequence, ε is the adaptive coefficient, ω i (t) is the noise sequence.

6. The method for predicting urban road traffic congestion index according to claim 4, characterized in that: The expressions of the first margin sequence r1(t) and the second characteristic mode function IMF2 are: r1(t)=x(t)-IMF1(t); 7. The method for predicting urban road traffic congestion index according to claim 4, characterized in that: The marginal sequence r of the k+1 stage and the kth stage k (t) and k+1 eigenmode components IMF k+1 , the expression is: r k (t)=r k-1 (t)-IMF k (t); 8. The method for predicting urban road traffic congestion index according to claim 4, characterized in that: The K intrinsic mode functions and the residual term expressions are:

9. The method for predicting urban road traffic congestion index according to claim 1, characterized in that: The training of each of the intrinsic mode function components comprises: By resetting the gate, the weight matrix W of Xr Get the output R of the reset gate t , the expression is: R t =σ(X t W xr +H t-1 W hr +b r ); Where R t represents the output of the reset gate at time t; X t Represents the current input; H t-1 represents the hidden state of the previous moment; δ represents the sigmoid function; W xr , W hr represents the weight parameter; b r represents the deviation parameter; By updating the gate, the hidden state H of the previous layer is t-1 Part of the data information is passed to the next layer for training, and the expression is: Z t =σ(X t W xz +H t-1 W hz +b z ) Where Z t represents the output of the update gate at time t; δ represents the sigmoid function; W xz , W hz represents the weight parameter; b z represents the deviation parameter; Use the output of the reset gate at time t to multiply the hidden state at time t-1 element by element, combine the calculated value with the input at time t, and get the candidate hidden state The expression is: in represents the candidate hidden state at time t; W xh , W hh represents the weight parameter; b h represents the bias parameter; ☉ represents element-wise multiplication; Based on the update gate Z t For H t-1 and Make a combination to get the hidden state H at time t t , the expression is: Among them, H t-1 Represents the hidden state of the previous moment; H t Represents the current hidden state; Represents the current candidate hidden state; Z t Represents the output of the current update gate; ☉ represents element-wise multiplication; Different weights are assigned according to the correlation of different positions in the input sequence. The expression is: Where Q represents the query vector; K represents the key vector; and V represents the value vector.

10. The method for predicting urban road traffic congestion index according to claim 1, characterized in that: The evaluation indicators include determination coefficient, mean absolute percentage error and mean absolute error.

Images