A method and system for evaluating the accuracy of highway transfer traffic volume prediction model

Abstract

The present invention discloses a method and system for evaluating the accuracy of a highway transfer traffic volume prediction model, including obtaining highway status data and historical data, constructing a highway data matching model based on the historical data, obtaining matching results from the highway data matching model, obtaining highway transfer traffic volume prediction values ​​and historical highway transfer traffic volume prediction values ​​through the highway transfer traffic volume prediction model, and determining a deviation index, determining the historical data prediction accuracy from the historical data, optimizing the highway data matching model based on the historical data prediction accuracy, determining the prediction model accuracy, and evaluating the accuracy of the highway transfer traffic volume prediction model. This method can not only improve the efficiency and accuracy of the accuracy evaluation of the highway transfer traffic volume prediction model, but also has good interpretability, and can be directly applied to the accuracy evaluation system of the highway transfer traffic volume prediction model.

Description

Technical Field

[0001] The present invention relates to the field of accuracy assessment, and in particular to an accuracy assessment method and system for a highway transfer traffic volume prediction model. Background Art

[0002] With the acceleration of urbanization and the rapid development of transportation infrastructure, the impact of road network construction on existing traffic flow is becoming increasingly significant. The resulting highway transfer is mainly predicted by the highway transfer traffic volume prediction model. Therefore, an effective accuracy evaluation of the prediction model is of great significance for traffic planning and management.

[0003] Traditional evaluation methods mostly focus on the simple error comparison between the predicted value of highway transfer traffic volume and the actual historical value, without considering the interaction of complex factors, resulting in the evaluation results cannot truly and comprehensively reflect the actual performance of the prediction model; at the same time, traditional evaluation methods also ignore the characteristic differences of different highway states in historical data. This single fixed evaluation method is difficult to adapt to diverse highway scenarios; in addition, the accuracy of the prediction model is also closely related to the number and quality of historical data used, the matching degree and correlation between historical data and current scenarios, and other factors. The lack of consideration of these factors in traditional evaluation methods is not conducive to the optimization and improvement of the prediction model. Therefore, based on the consideration of the above evaluation factors, an accuracy evaluation of the highway transfer traffic volume prediction model is designed to overcome the shortcomings of the existing accuracy evaluation methods, accurately measure the performance of the model, and provide strong support for scientific decision-making in the transportation field and the continuous optimization of the prediction model. Summary of the invention

[0004] The purpose of the present invention is to provide a method and system for evaluating the accuracy of a highway transfer traffic volume prediction model.

[0005] To achieve the above object, the present invention is implemented according to the following technical solutions:

[0006] The present invention comprises the following steps:

[0007] Obtaining highway status data and historical data, constructing a highway data matching model based on the historical data, and inputting the highway status data into the highway data matching model to obtain a matching result; the historical data includes historical highway status data and historical highway transfer traffic volume; the matching result includes matching historical data, similarity index, correlation coefficient and number of matching data groups;

[0008] Obtaining a highway transfer traffic volume prediction value and a historical highway transfer traffic volume prediction value through a highway transfer traffic volume prediction model, and determining a deviation index according to the highway transfer traffic volume prediction value and the historical highway transfer traffic volume prediction value;

[0009] Determining historical data prediction accuracy from the historical data, determining a first accuracy and a second accuracy according to the historical data prediction accuracy and the similarity index, and optimizing the highway data matching model according to the first accuracy and the second accuracy;

[0010] Determine a third accuracy according to the first accuracy, the deviation index, the correlation coefficient and the number of matching data groups, determine a fourth accuracy according to the second accuracy and the deviation index, and optimize the highway data matching model according to the third accuracy and the fourth accuracy;

[0011] The prediction model accuracy is determined according to the third accuracy and the fourth accuracy, and a prediction model accuracy threshold is determined to evaluate the accuracy of the highway transfer traffic volume prediction model.

[0012] Furthermore, the method for obtaining the matching result includes:

[0013] Obtain highway status data and historical data, pre-process the highway status data and historical data, and use the historical data to build a highway data matching model;

[0014] The highway data matching model includes a data matching module, a correlation coefficient module and a loss function module;

[0015] The data matching module is used to calculate the similarity between the highway state data and the historical highway state data, determine the matched historical highway state data, similarity index and the number of matched data groups according to the similarity calculation result, and determine the corresponding historical highway transfer traffic volume according to the matched historical highway state data; the similarity index includes average similarity, maximum similarity, minimum similarity and similarity coefficient, the average similarity is determined by the mean of the similarity between the highway state data and all the matched historical highway state data, and the similarity coefficient is determined by the maximum similarity, minimum similarity and average similarity between the highway state data and the matched historical highway state data;

[0016] The correlation coefficient module is used to calculate the correlation between the historical highway status data and the historical highway transfer traffic volume in the matching historical data, and the average correlation between the historical highway status data and all the matched historical data is used as the correlation coefficient;

[0017] The loss function module includes a first loss function and a second loss function; the first loss function is used to optimize the highway data matching model according to the first accuracy and the second accuracy; the second loss function is used to optimize the highway data matching model according to the third accuracy and the fourth accuracy;

[0018] The road state data to be matched is input into the road data matching model to obtain a matching result.

[0019] Furthermore, the method for determining the deviation index comprises the following steps:

[0020] Based on the matching results, the highway status data is input into the highway transfer traffic volume prediction model to obtain the highway transfer traffic volume prediction value, and the historical highway status data is input into the highway transfer traffic volume prediction model to obtain the historical highway transfer traffic volume prediction value;

[0021] The deviation index includes average data deviation, maximum data deviation, minimum data deviation and deviation coefficient. The data deviation between the highway transfer traffic volume forecast value and all historical highway transfer traffic volume forecast values ​​is calculated, and the deviation coefficient is determined based on the average data deviation, maximum data deviation and minimum data deviation.

[0022] Furthermore, the method for optimizing the highway data matching model according to the first accuracy and the second accuracy includes:

[0023] The historical data prediction accuracy is obtained by averaging the accuracy of all historical highway transfer volumes and historical highway transfer volume prediction value data sets;

[0024] Determining a first accuracy by the historical data prediction accuracy and the average similarity; the first accuracy is positively correlated with the historical data prediction accuracy and the average similarity;

[0025] The second accuracy is determined by the historical data prediction accuracy and the similarity coefficient; the second accuracy is positively correlated with the historical data prediction accuracy and the similarity coefficient;

[0026] When the first accuracy is less than the second accuracy, the highway data matching model is optimized by the first loss function, and the first accuracy and the second accuracy are calculated by rematching the historical data according to the optimized highway data matching model, and the above operation is repeated until the first accuracy is greater than or equal to the second accuracy and the optimization is stopped; the first loss function expression is:

[0027]

[0028] in is the first loss function, is the matrix weight, is the category weight, is the number of matching data sets, Highway status data and The similarity matrix of group matching historical highway status data, For the The group matches the labels of historical highway status data. is the similarity between two vectors, is the highway state vector, To match the historical highway status average vector, Traffic flow highway state vector, For the Group traffic flow classes to match historical highway state vectors, is the state vector of the social and economic highway, For the Group socioeconomic classes to match historical highway state vectors, is the highway state vector of the road network class, For the Group road network classes to match historical highway state vectors, Update the period hyperparameters for the database, is the first regularization strength, for Regularization term, For the first accuracy, The second accuracy.

[0029] Further, the method for determining the third accuracy includes:

[0030] Determining a third accuracy according to the first accuracy, the average data deviation, the correlation coefficient and the number of matching data groups; the third accuracy is positively correlated with the first accuracy, the correlation coefficient and the number of matching data groups; the third accuracy is negatively correlated with the average data deviation;

[0031] The third accuracy expression is:

[0032]

[0033] in For the third degree of accuracy, For the first accuracy, is the bias weight, is the average data deviation, is the relevance weight, is the correlation coefficient, is the periodic change angle parameter of the historical database, is the group weight, is the number of matching data sets.

[0034] Furthermore, the method for optimizing the highway data matching model according to the third accuracy and the fourth accuracy includes:

[0035] Determining a fourth accuracy from the second accuracy and the coefficient of deviation; the fourth accuracy is positively correlated with the second accuracy; the fourth accuracy is negatively correlated with the coefficient of deviation;

[0036] When the third accuracy is less than the fourth accuracy, the highway data matching model is optimized by the second loss function, and the third accuracy and the fourth accuracy are calculated by rematching the historical data according to the optimized highway data matching model, and the above operation is repeated until the third accuracy is greater than or equal to the fourth accuracy and the optimization is stopped; the second loss function expression is:

[0037]

[0038] in is the second loss function, is the second regularization strength, For the third degree of accuracy, For the fourth degree of accuracy, is the strength of correlation, is the group strength.

[0039] Furthermore, the method for evaluating the accuracy of the highway transfer traffic volume prediction model includes:

[0040] The third accuracy and the fourth accuracy are subjected to data fusion to obtain a fifth accuracy, and the accuracy of the prediction model is determined by the third accuracy and the fifth accuracy;

[0041] The specific steps of the data fusion are to fuse the third accuracy and the fourth accuracy by using weighted average method, Bayesian method, data mining and machine learning respectively, and perform secondary average weighting on the data fusion results of the three methods to obtain the fifth accuracy;

[0042] According to the second accuracy, the fourth accuracy, the similarity coefficient and the deviation coefficient, an accuracy threshold search is performed to obtain the prediction model accuracy threshold. When the prediction model accuracy is greater than the prediction model accuracy threshold, it is determined that the highway transfer traffic volume prediction model has a high accuracy. Otherwise, it is determined that the highway transfer traffic volume prediction model has insufficient accuracy and there is room for optimization.

[0043] In a second aspect, a highway transfer traffic volume prediction model accuracy evaluation system includes:

[0044] Data acquisition module: used to obtain highway status data and historical data, and pre-process the highway status data and historical data;

[0045] Data matching module: used for constructing a highway data matching model according to the historical data, inputting the highway state data into the highway data matching model to obtain a matching result, and optimizing the highway data matching model according to the first loss function and the second loss function;

[0046] Prediction model module: used to obtain the highway transfer traffic volume prediction value and the historical highway transfer traffic volume prediction value according to the highway transfer traffic volume prediction model;

[0047] Accuracy module: used to determine the deviation index according to the highway transfer traffic volume prediction value and the historical highway transfer traffic volume prediction value, used to determine the first accuracy, the second accuracy, the third accuracy, the fourth accuracy and the prediction model accuracy, used to perform accuracy threshold search to obtain the prediction model accuracy threshold, and used to perform accuracy evaluation according to the prediction model accuracy and the prediction model accuracy threshold;

[0048] Intelligent supervision module: used to store, view and manage the highway status data, the historical data and the accuracy of the prediction model, and to prompt the user according to the accuracy evaluation result of the highway transfer traffic volume prediction model.

[0049] The beneficial effects of the present invention are:

[0050] The present invention is a method and system for evaluating the accuracy of a highway transfer traffic volume prediction model. Compared with the prior art, the present invention has the following technical effects:

[0051] The present invention can improve the data preprocessing capabilities and enhance the model adaptability in the accuracy assessment of the highway transfer traffic volume prediction model through data matching, determining deviation indicators, determining the prediction accuracy of historical data, optimizing models, determining the accuracy of the prediction model and threshold search steps, thereby improving the efficiency and precision of the accuracy assessment of the highway transfer traffic volume prediction model. The accuracy assessment technology of the highway transfer traffic volume prediction model is optimized, which can greatly save resources and improve work efficiency, realize the evaluation of the accuracy of the highway transfer traffic volume prediction model, and provide data reference for the optimization of the highway transfer traffic volume prediction model. It is of great significance to traffic planning and management, and can adapt to the terminal traceability needs of the whole industry chain quality traceability systems of different insulation projects and the whole industry chain quality traceability of insulation projects of different users, and has a certain universality. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 The present invention is a flowchart of the steps of a method for evaluating the accuracy of a highway transfer traffic volume prediction model. DETAILED DESCRIPTION

[0053] The present invention is further described below by means of specific embodiments. The illustrative embodiments and descriptions of the present invention are used to explain the present invention but are not intended to limit the present invention.

[0054] The present invention provides a method and system for evaluating the accuracy of a highway transfer traffic volume prediction model, comprising the following steps:

[0055] like Figure 1 As shown, in this embodiment, the following steps are included:

[0056] Obtaining highway status data and historical data, constructing a highway data matching model based on the historical data, and inputting the highway status data into the highway data matching model to obtain a matching result; the historical data includes historical highway status data and historical highway transfer traffic volume; the matching result includes matching historical data, similarity index, correlation coefficient and number of matching data groups;

[0057] Obtaining a highway transfer traffic volume prediction value and a historical highway transfer traffic volume prediction value through a highway transfer traffic volume prediction model, and determining a deviation index according to the highway transfer traffic volume prediction value and the historical highway transfer traffic volume prediction value;

[0058] Determining historical data prediction accuracy from the historical data, determining a first accuracy and a second accuracy according to the historical data prediction accuracy and the similarity index, and optimizing the highway data matching model according to the first accuracy and the second accuracy;

[0059] Determine a third accuracy according to the first accuracy, the deviation index, the correlation coefficient and the number of matching data groups, determine a fourth accuracy according to the second accuracy and the deviation index, and optimize the highway data matching model according to the third accuracy and the fourth accuracy;

[0060] The prediction model accuracy is determined according to the third accuracy and the fourth accuracy, and a prediction model accuracy threshold is determined to evaluate the accuracy of the highway transfer traffic volume prediction model.

[0061] In this embodiment, the method for obtaining the matching result includes:

[0062] Obtain highway status data and historical data; the highway status data includes traffic flow data, social and economic data, and road network data; the traffic flow data includes traffic flow, road speed, congestion time, traffic accidents / controls / construction conditions, and the resulting traffic transfer; the social and economic data includes total population, population density, economic level, income level, food production, land type, and land area; the road network data includes road section data, node data, and traffic facility data;

[0063] Preprocessing the historical data, and using the historical data to build a highway data matching model; the highway data matching model includes a data matching module, a correlation coefficient module and a loss function module;

[0064] The data matching module is used to calculate the similarity between the highway state data and the historical highway state data, determine the matched historical highway state data, similarity index and the number of matched data groups according to the similarity calculation result, and determine the corresponding historical highway transfer traffic volume according to the matched historical highway state data; the similarity index includes average similarity, maximum similarity, minimum similarity and similarity coefficient, the average similarity is determined by the mean of the similarity between the highway state data and all the matched historical highway state data, and the similarity coefficient is determined by the maximum similarity, minimum similarity and average similarity between the highway state data and the matched historical highway state data;

[0065] The correlation coefficient module is used to calculate the correlation between the historical highway status data and the historical highway transfer traffic volume in the matching historical data, and the average correlation between the historical highway status data and all the matched historical data is used as the correlation coefficient;

[0066] The loss function module includes a first loss function and a second loss function; the first loss function is used to optimize the highway data matching model according to the first accuracy and the second accuracy; the second loss function is used to optimize the highway data matching model according to the third accuracy and the fourth accuracy;

[0067] Inputting the to-be-matched highway state data into the highway data matching model to obtain a matching result;

[0068] In the actual evaluation, the historical highway state data matched by the highway state data corresponds to the historical highway traffic transfer events whose traffic flow conditions, socioeconomic conditions and road network conditions are highly similar to the current highway state. By matching the historical highway state data with high similarity, it is ensured that the highway traffic transfer events input into the prediction model correspond to the same highway state. At the same time, a set of highway state data needs to match at least 5 sets of historical highway state data. For example, the similarities between a certain highway state data and the historical highway state data are 0.98, 0.95, 0.92, 0.91, 0.9, 0.88, 0.88, 0.86, 0.84, 0.82 and 0.80 respectively. If the screening condition is set to a similarity greater than or equal to 0.85, the 8 sets of historical highway state data that meet the screening condition are determined to be the matching historical highway state data. If the screening condition is set to a similarity greater than or equal to 0.91, at this time, the similarity of only 4 sets of data meets the screening condition. The screening condition is adjusted to a similarity greater than or equal to 0.9, so that at least 5 sets of historical highway state data are matched.

[0069] Each set of historical highway status data has a unique corresponding historical highway transfer traffic volume. The historical highway transfer traffic volume refers to other highway routes to which traffic flow is transferred, the transfer ratio of traffic flow to other highway routes, the specific types of traffic flow in other highway routes, other modes of transportation to which traffic flow is transferred, the transfer ratio of traffic flow to other modes of transportation, and the specific types of traffic flow in other modes of transportation in highway traffic transfer events recorded in real historical records. The specific types of traffic flow include freight flow, passenger flow, public transportation flow, work vehicle flow, non-motor vehicle flow and pedestrian flow.

[0070] In this embodiment, the method for determining the deviation index includes the following steps:

[0071] Based on the matching results, the highway status data is input into the highway transfer traffic volume prediction model to obtain the highway transfer traffic volume prediction value, and the historical highway status data is input into the highway transfer traffic volume prediction model to obtain the historical highway transfer traffic volume prediction value;

[0072] The deviation index includes average data deviation, maximum data deviation, minimum data deviation and deviation coefficient. The data deviation between the predicted value of highway transfer traffic volume and all historical highway transfer traffic volume predicted values ​​is calculated. The deviation coefficient is determined according to the average data deviation, maximum data deviation and minimum data deviation. The expression is:

[0073]

[0074] in is the coefficient of variation, is the average data deviation, is the maximum data deviation, is the minimum data deviation;

[0075] In the actual evaluation, the matching screening condition of the matching model is set to a similarity greater than or equal to 0.85, and 8 sets of matching historical highway status data are obtained. The highway status data and the 8 sets of historical highway status data are input into the highway transfer traffic volume prediction model to obtain the highway transfer traffic volume prediction value and the 8 sets of historical highway transfer traffic volume prediction values;

[0076] The deviation index of the predicted value is calculated respectively according to the transferred traffic volume of other highway routes and the transferred traffic volume of other modes of transportation. The average weighted deviation of the data group corresponding to the maximum deviation of the transferred traffic volume of other highway routes is taken as the maximum data deviation, the average weighted deviation of the data group corresponding to the minimum deviation of the transferred traffic volume of other modes of transportation is taken as the minimum data deviation, and the average weighted deviation of all data groups is taken as the average data deviation; the average weighted deviation is the mean of all highway transfer traffic volume indicators; the maximum data deviation of the predicted value of highway transfer traffic volume calculated according to the above method is 0.23, the minimum data deviation is 0.11, the average data deviation is 0.15, and the calculated deviation coefficient is 0.164.

[0077] In this embodiment, the method for optimizing the highway data matching model according to the first accuracy and the second accuracy includes:

[0078] The historical data prediction accuracy is obtained by averaging the accuracy of all historical highway transfer volumes and historical highway transfer volume prediction value data sets;

[0079] Determining a first accuracy by the historical data prediction accuracy and the average similarity; the first accuracy is positively correlated with the historical data prediction accuracy and the average similarity;

[0080] The second accuracy is determined by the historical data prediction accuracy and the similarity coefficient; the second accuracy is positively correlated with the historical data prediction accuracy and the similarity coefficient;

[0081] When the first accuracy is less than the second accuracy, the highway data matching model is optimized by the first loss function, and the first accuracy and the second accuracy are calculated by rematching the historical data according to the optimized highway data matching model, and the above operation is repeated until the first accuracy is greater than or equal to the second accuracy and the optimization is stopped; the first loss function expression is:

[0082]

[0083] in is the first loss function, is the matrix weight, is the category weight, is the number of matching data sets, Highway status data and The similarity matrix of group matching historical highway status data, For the The group matches the labels of historical highway status data. is the similarity between two vectors, is the highway state vector, To match the historical highway status average vector, Traffic flow highway state vector, For the Group traffic flow classes to match historical highway state vectors, is the state vector of the social and economic highway, For the Group socioeconomic classes to match historical highway state vectors, is the highway state vector of the road network class, For the Group road network classes to match historical highway state vectors, Update the period hyperparameters for the database, is the first regularization strength, for Regularization term, For the first accuracy, is the second accuracy;

[0084] In the actual evaluation, the average similarity of 8 sets of matching data is , the maximum similarity , minimum similarity ,according to Calculate the similarity coefficient ; According to the historical data prediction accuracy, take 0.93, and take the first accuracy according to the product of the average similarity and the historical data prediction accuracy , take the second accuracy by multiplying the similarity coefficient by the historical data prediction accuracy , at this time the first accuracy Less than second accuracy ;

[0085] Get the matrix weight , Category weight 2. First regularization strength , optimize the highway data matching model according to the first loss function and recalculate the first accuracy , second accuracy , at this time the first accuracy Greater than second accuracy , and proceed to the next calculation step.

[0086] In this embodiment, the method for determining the third accuracy includes:

[0087] Determining a third accuracy according to the first accuracy, the average data deviation, the correlation coefficient and the number of matching data groups; the third accuracy is positively correlated with the first accuracy, the correlation coefficient and the number of matching data groups; the third accuracy is negatively correlated with the average data deviation;

[0088] The third accuracy expression is:

[0089]

[0090] in For the third degree of accuracy, For the first accuracy, is the bias weight, is the average data deviation, is the relevance weight, is the correlation coefficient, is the periodic change angle parameter of the historical database, is the group weight, is the number of matching data groups;

[0091] In the actual evaluation, the highway data matching model optimized by the first loss function matched 6 sets of historical highway status data, and the corresponding average data deviation , correlation coefficient , Historical database periodic change angle parameters , bias weight , relevance weight , Group weight , the third accuracy is calculated 0.962.

[0092] In this embodiment, the method for optimizing the highway data matching model according to the third accuracy and the fourth accuracy includes:

[0093] Determining a fourth accuracy from the second accuracy and the coefficient of deviation; the fourth accuracy is positively correlated with the second accuracy; the fourth accuracy is negatively correlated with the coefficient of deviation;

[0094] When the third accuracy is less than the fourth accuracy, the highway data matching model is optimized by the second loss function, and the third accuracy and the fourth accuracy are calculated by rematching the historical data according to the optimized highway data matching model, and the above operation is repeated until the third accuracy is greater than or equal to the fourth accuracy and the optimization is stopped; the second loss function expression is:

[0095]

[0096] in is the second loss function, is the second regularization strength, For the third degree of accuracy, For the fourth degree of accuracy, is the strength of correlation, is the group strength;

[0097] In the actual evaluation, the highway data matching model optimized according to the first loss function matched 6 sets of historical highway status data, with the corresponding maximum data deviation of 0.21, the minimum data deviation of 0.11, the average data deviation of 0.13, and the calculated deviation coefficient of 0.118;

[0098] Pick ,according to Calculate the fourth accuracy , at this time, the third accuracy is greater than the fourth accuracy, and the next calculation step can be directly entered; if the third accuracy is less than the fourth accuracy, the highway data matching model needs to be optimized according to the second loss function, and the second regularization strength is , repeat the above operation until the third accuracy is greater than or equal to the fourth accuracy, and then proceed to the next calculation step.

[0099] In this embodiment, the method for evaluating the accuracy of the highway transfer traffic volume prediction model includes:

[0100] The third accuracy and the fourth accuracy are subjected to data fusion to obtain a fifth accuracy, and the accuracy of the prediction model is determined by the third accuracy and the fifth accuracy;

[0101] The specific steps of the data fusion are to fuse the third accuracy and the fourth accuracy by using the weighted average method, the Bayesian method, and data mining and machine learning respectively, and perform quadratic average weighting on the data fusion results of the three methods to obtain the fifth accuracy; the weighted average method assigns different weights to the third accuracy and the fourth accuracy for weighted summation; the Bayesian method calculates the posterior probability of the third accuracy and the fourth accuracy based on the probability theory method for weighted summation; the data mining and machine learning fusion adopts a linear regression algorithm to learn the relationship between the accuracies and predict the data fusion results;

[0102] Performing an accuracy threshold search according to the second accuracy, the fourth accuracy, the similarity coefficient and the deviation coefficient to obtain a prediction model accuracy threshold, and when the prediction model accuracy is greater than the prediction model accuracy threshold, it is determined that the highway transfer traffic volume prediction model has a high accuracy; otherwise, it is determined that the highway transfer traffic volume prediction model has insufficient accuracy and there is room for optimization;

[0103] In the actual evaluation, the third accuracy and the fourth accuracy were 0.962 and 0.871 respectively. The weighted average method, Bayesian method, data mining and machine learning were used to fuse the third accuracy and the fourth accuracy respectively, and the fusion results were 0.9165, 0.9438 and 0.9339. The fifth accuracy was 0.9314 after the second average weighting. The accuracy of the prediction model was determined to be 0.9467 based on the third accuracy and the fifth accuracy.

[0104] The accuracy threshold is searched according to the second accuracy, the fourth accuracy, the similarity coefficient and the deviation coefficient to obtain the accuracy threshold of the prediction model. The specific steps are:

[0105] Set the initial threshold and calculate the fitness of the initial threshold:

[0106]

[0107] in is the fitness of the initial threshold, is the similarity coefficient weight, is the deviation coefficient weight;

[0108] According to the prediction model accuracy threshold, it is determined whether to update the position and velocity of the accuracy particle. The expression is:

[0109]

[0110]

[0111] in For the The accuracy of the particle is The speed at the iteration, For the The accuracy of the particle is The speed at the iteration, is the accuracy factor, , for A random number, , is the acceleration constant, For the The accuracy of the particle is The position at the iteration, No. The accuracy of the particle is The historical best position at the iteration, For the The historical optimal position of all accuracy particles at the iteration, For the The accuracy of the particle is The position at the iteration;

[0112] According to the updated position and speed of the accuracy particle, the fitness of the current prediction model accuracy threshold is recalculated, and the above operation is repeated until the fitness of the current prediction model accuracy threshold is the optimal fitness or the maximum number of iterations is reached, then the iteration is stopped and the accuracy threshold is output;

[0113] After searching, the output accuracy threshold is 0.9200, and the prediction model accuracy of 0.9467 is greater than the accuracy threshold of 0.9200, which means that the highway transfer traffic volume prediction model has a high accuracy.

[0114] In a second aspect, a highway transfer traffic volume prediction model accuracy evaluation system includes:

[0115] Data acquisition module: used to obtain highway status data and historical data, and pre-process the highway status data and historical data;

[0116] Data matching module: used for constructing a highway data matching model according to the historical data, inputting the highway state data into the highway data matching model to obtain a matching result, and optimizing the highway data matching model according to the first loss function and the second loss function;

[0117] Prediction model module: used to obtain the highway transfer traffic volume prediction value and the historical highway transfer traffic volume prediction value according to the highway transfer traffic volume prediction model;

[0118] Accuracy module: used to determine the deviation index according to the highway transfer traffic volume prediction value and the historical highway transfer traffic volume prediction value, used to determine the first accuracy, the second accuracy, the third accuracy, the fourth accuracy and the prediction model accuracy, used to perform accuracy threshold search to obtain the prediction model accuracy threshold, and used to perform accuracy evaluation according to the prediction model accuracy and the prediction model accuracy threshold;

[0119] Intelligent supervision module: used to store, view and manage the highway status data, the historical data and the accuracy of the prediction model, and to prompt the user according to the accuracy evaluation result of the highway transfer traffic volume prediction model.

[0120] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for evaluating the accuracy of a highway transfer traffic volume prediction model, characterized in that: The following steps are involved: S1, obtaining highway status data and historical data, constructing a highway data matching model according to the historical data, and inputting the highway status data into the highway data matching model to obtain a matching result; the historical data includes historical highway status data and historical highway transfer traffic volume; the matching result includes matching historical data, similarity index, correlation coefficient and number of matching data groups; S2. Obtaining a highway transfer traffic volume prediction value and a historical highway transfer traffic volume prediction value through a highway transfer traffic volume prediction model, and determining a deviation index according to the highway transfer traffic volume prediction value and the historical highway transfer traffic volume prediction value; The method for determining the deviation index comprises the following steps: Based on the matching results, the highway status data is input into the highway transfer traffic volume prediction model to obtain the highway transfer traffic volume prediction value, and the historical highway status data is input into the highway transfer traffic volume prediction model to obtain the historical highway transfer traffic volume prediction value; The deviation index includes average data deviation, maximum data deviation, minimum data deviation and deviation coefficient, and the data deviation between the predicted value of highway transfer traffic volume and all historical highway transfer traffic volume predicted values ​​is calculated, and the deviation coefficient is determined according to the average data deviation, maximum data deviation and minimum data deviation; S3, determining the historical data prediction accuracy from the historical data, determining a first accuracy and a second accuracy according to the historical data prediction accuracy and the similarity index, and optimizing the highway data matching model according to the first accuracy and the second accuracy; S4, determining a third accuracy according to the first accuracy, the deviation index, the correlation coefficient and the number of matching data groups, and determining a fourth accuracy according to the second accuracy and the deviation index; When the third accuracy is less than the fourth accuracy, the highway data matching model is optimized by the second loss function, and the third accuracy and the fourth accuracy are calculated by rematching the historical data according to the optimized highway data matching model, and the above operation is repeated until the third accuracy is greater than or equal to the fourth accuracy and the optimization is stopped; the second loss function expression is: in is the second loss function, is the second regularization strength, For the third degree of accuracy, For the fourth degree of accuracy, is the strength of correlation, The group strength optimizing the highway data matching model according to the third accuracy and the fourth accuracy; S5. Determine the accuracy of the prediction model according to the third accuracy and the fourth accuracy, and determine a prediction model accuracy threshold to evaluate the accuracy of the highway transfer traffic volume prediction model.

2. The accuracy evaluation method of a highway transfer traffic volume prediction model according to claim 1 is characterized in that: The method for obtaining the matching result includes: Obtain highway status data and historical data, pre-process the highway status data and historical data, and use the historical data to build a highway data matching model; The highway data matching model includes a data matching module, a correlation coefficient module and a loss function module; The data matching module is used to calculate the similarity between the highway state data and the historical highway state data, determine the matched historical highway state data, similarity index and the number of matched data groups according to the similarity calculation result, and determine the corresponding historical highway transfer traffic volume according to the matched historical highway state data; the similarity index includes average similarity, maximum similarity, minimum similarity and similarity coefficient, the average similarity is determined by the mean of the similarity between the highway state data and all the matched historical highway state data, and the similarity coefficient is determined by the maximum similarity, minimum similarity and average similarity between the highway state data and the matched historical highway state data; The correlation coefficient module is used to calculate the correlation between the historical highway status data and the historical highway transfer traffic volume in the matching historical data, and the average correlation between the historical highway status data and all the matched historical data is used as the correlation coefficient; The loss function includes a first loss function and a second loss function; the first loss function is used to optimize the highway data matching model according to the first accuracy and the second accuracy; the second loss function is used to optimize the highway data matching model according to the third accuracy and the fourth accuracy; The road state data to be matched is input into the road data matching model to obtain a matching result.

3. The accuracy evaluation method of a highway transfer traffic volume prediction model according to claim 1 is characterized in that: The method for optimizing the highway data matching model according to the first accuracy and the second accuracy includes: The historical data prediction accuracy is obtained by averaging the accuracy of all historical highway transfer volumes and historical highway transfer volume prediction value data sets; Determining a first accuracy by the historical data prediction accuracy and the average similarity; the first accuracy is positively correlated with the historical data prediction accuracy and the average similarity; The second accuracy is determined by the historical data prediction accuracy and the similarity coefficient; the second accuracy is positively correlated with the historical data prediction accuracy and the similarity coefficient; When the first accuracy is less than the second accuracy, the highway data matching model is optimized by the first loss function, and the historical data is re-matched according to the optimized highway data matching model to calculate the first accuracy and the second accuracy, and the above operation is repeated until the first accuracy is greater than or equal to the second accuracy and the optimization is stopped; the first loss function expression is: in is the first loss function, is the matrix weight, is the category weight, is the number of matching data sets, Highway status data and The similarity matrix of group matching historical highway status data, For the The group matches the labels of historical highway status data. is the similarity between two vectors, is the highway state vector, To match the historical highway status average vector, Traffic flow highway state vector, For the Group traffic flow classes to match historical highway state vectors, is the state vector of the social and economic highway, For the Group socioeconomic classes to match historical highway state vectors, is the highway state vector of the road network class, For the Group road network classes to match historical highway state vectors, Update the period hyperparameters for the database, is the first regularization strength, for Regularization term, For the first accuracy, The second accuracy.

4. The accuracy evaluation method of a highway transfer traffic volume prediction model according to claim 1 is characterized in that: The method of determining the third accuracy comprises: Determining a third accuracy according to the first accuracy, the average data deviation, the correlation coefficient and the number of matching data groups; the third accuracy is positively correlated with the first accuracy, the correlation coefficient and the number of matching data groups; the third accuracy is negatively correlated with the average data deviation; The third accuracy expression is: in For the third degree of accuracy, For the first accuracy, is the bias weight, is the average data deviation, is the relevance weight, is the correlation coefficient, is the periodic change angle parameter of the historical database, is the group weight, is the number of matching data sets.

5. The accuracy evaluation method of a highway transfer traffic volume prediction model according to claim 1 is characterized in that: The method for optimizing the highway data matching model according to the third accuracy and the fourth accuracy includes: The fourth accuracy is determined by the second accuracy and the deviation coefficient; the fourth accuracy is positively correlated with the second accuracy; and the fourth accuracy is negatively correlated with the deviation coefficient.

6. The accuracy evaluation method of a highway transfer traffic volume prediction model according to claim 1 is characterized in that: The method for evaluating the accuracy of the highway transfer traffic volume prediction model comprises: The third accuracy and the fourth accuracy are subjected to data fusion to obtain a fifth accuracy, and the accuracy of the prediction model is determined by the third accuracy and the fifth accuracy; The specific steps of the data fusion are to fuse the third accuracy and the fourth accuracy by using weighted average method, Bayesian method, data mining and machine learning respectively, and perform secondary average weighting on the data fusion results of the three methods to obtain the fifth accuracy; According to the second accuracy, the fourth accuracy, the similarity coefficient and the deviation coefficient, an accuracy threshold search is performed to obtain the prediction model accuracy threshold. When the prediction model accuracy is greater than the prediction model accuracy threshold, it is determined that the highway transfer traffic volume prediction model has a high accuracy. Otherwise, it is determined that the highway transfer traffic volume prediction model has insufficient accuracy and there is room for optimization.

7. A highway transfer traffic volume prediction model accuracy evaluation system, used to execute the method according to any one of claims 1 to 6, characterized in that: include: Data acquisition module: used to obtain highway status data and historical data, and pre-process the highway status data and historical data; Data matching module: used to construct a highway data matching model according to the historical data, input the highway status data into the highway data matching model to obtain a matching result, and optimize the highway data matching model according to the loss function; Prediction model module: used to obtain the highway transfer traffic volume prediction value and the historical highway transfer traffic volume prediction value according to the highway transfer traffic volume prediction model; Accuracy module: used to determine the deviation index according to the highway transfer traffic volume prediction value and the historical highway transfer traffic volume prediction value, used to determine the first accuracy, the second accuracy, the third accuracy, the fourth accuracy and the prediction model accuracy, used to perform accuracy threshold search to obtain the prediction model accuracy threshold, and used to perform accuracy evaluation according to the prediction model accuracy and the prediction model accuracy threshold; Intelligent supervision module: used to store, view and manage the highway status data, the historical data and the accuracy of the prediction model, and to prompt the user according to the accuracy evaluation result of the highway transfer traffic volume prediction model.

Images