TWI676397B - Artificial intelligence traffic estimation system using mobile network signaling data and method thereof - Google Patents

Abstract

An artificial intelligence vehicle flow estimation system and method using mobile network signaling data, the system includes a mobile network signaling data acquisition module, an unsupervised learning module, a feature extraction module, and a supervisor Learning module. The mobile network signaling data extraction module captures mobile network signaling data between a plurality of mobile devices and a network within a predetermined range of a specified road; the unsupervised learning module converts the mobile network according to the speed of the mobile device The channel signaling data is divided into different ethnic groups; the feature extraction module extracts specific ethnic groups from different ethnic groups, and calculates the amount of mobile network signaling data per unit time as the extracted characteristics; the supervised learning module extracts The latter feature establishes an artificial intelligence traffic flow estimation model to estimate the traffic flow information of a plurality of vehicles passing a designated road.

Description

Artificial intelligence traffic flow estimation system and method using mobile network signaling data

The invention relates to a traffic flow estimation technology, and particularly to an artificial intelligence traffic flow estimation system and method using mobile network signaling data.

In recent years, compared with traditional fixed vehicle detectors (VD) and road toll detection (ETC) based on the detection of road traffic information, ETC-Based Vehicle Probe (EVP) and GPS-Based Vehicle Probe (GVP), etc. Cellular-Based Vehicle Probe (CVP) technology based on mobile device base stations has a wide coverage and low cost It is not necessary to use the communication device on the vehicle. It has become one of the hot research topics.

In the past research, CVP signaling data was mainly used to estimate vehicle speed or travel time. The reason is that the average speed and travel time of a road segment can be calculated from the time difference, location difference, and length of a section of CVP signaling data. However, if the number of CVP signaling data on the road is used as the vehicle The traffic often produces non-negligible errors, because a vehicle usually has more than one passenger, that is, the amount of CVP signaling data generated by each vehicle is not the same. Therefore, on dense roads, the total number of CVP signaling data is much higher than the total number of vehicles, and on general roads, the traffic flow pattern is a mixed flow of steam and motorcycles. The distribution pattern of CVP signaling data will vary with There is a significant difference in the proportion of roads and motorcycles or the shape of roads and roads, which makes it difficult to estimate the traffic volume.

Therefore, how to solve the above-mentioned shortcomings of the prior art has become a major issue for those skilled in the art.

The invention provides an artificial intelligence traffic flow estimation system and method using mobile network signaling data, which can use mobile network signaling data to estimate traffic flow information of a plurality of vehicles passing a designated road.

The artificial intelligence vehicle flow estimation system using mobile network signaling data in the present invention includes: a mobile network signaling data acquisition module for capturing a plurality of mobile devices on a specified road through a specified road A plurality of mobile network signaling data between at least one network; an unsupervised learning module, which is an unsupervised learning algorithm with artificial intelligence to use the unsupervised learning algorithms to rely on the mobile devices Speed, the mobile network signaling data extraction module is divided into a plurality of different ethnic groups; a feature extraction module, from the unsupervised learning algorithm divided into different At least one specific group is extracted from the ethnic group, and the number of mobile network signaling data per unit time in the at least one specific group is calculated as the extracted characteristics; and a supervised learning module having One of the artificial intelligence supervised learning algorithms, using the supervised learning algorithm to build an artificial intelligence traffic flow estimation model from the features extracted by the feature extraction module, and then use the artificial intelligence traffic flow estimation model to estimate or generate a pass Traffic information for multiple vehicles on a given road.

The artificial intelligence traffic flow estimation method using mobile network signaling data in the present invention includes: capturing a plurality of mobile network signaling between a plurality of mobile devices passing through a designated road and at least one network within a predetermined range of a designated road Data; using an unsupervised learning algorithm with artificial intelligence to divide the mobile network signaling data into a number of different ethnic groups based on the speed of the mobile devices; the differences from the unsupervised learning algorithm At least one specific group is extracted from the ethnic group, and the number of mobile network signaling data per unit time in the at least one specific group is calculated as the extracted characteristics; and a supervised learning algorithm with artificial intelligence is used, The feature extracted by the feature extraction module establishes an artificial intelligence traffic flow estimation model to estimate or generate traffic flow information of a plurality of vehicles passing a designated road through the artificial intelligence traffic flow estimation model.

In order to make the above features and advantages of the present invention more comprehensible, embodiments are described below in detail with reference to the accompanying drawings. Additional features and advantages of the present invention will be partially explained in the following description, and these features and advantages will be partially obvious from the description, or may be learned through practice of the present invention. The features and advantages of the invention are realized and achieved by means of elements and combinations specifically pointed out in the scope of the patent application. It should be understood that both the foregoing general description and the following detailed description are merely exemplary and explanatory and are not intended to limit the scope of the invention as claimed.

1‧‧‧Artificial intelligence traffic flow estimation system

10‧‧‧Mobile Network Signaling Data Retrieval Module

20‧‧‧Unsupervised Learning Module

21‧‧‧Unsupervised Learning Algorithm

30‧‧‧Feature Extraction Module

40‧‧‧Supervised Learning Module

41‧‧‧Supervised Learning Algorithm

50‧‧‧ Artificial Intelligence Traffic Flow Estimation Model

51‧‧‧input layer

52‧‧‧LSTM (Long Short-Term Memory) Layer

53‧‧‧Hidden layer

54‧‧‧Output layer

60‧‧‧Traffic Information

A‧‧‧mobile device

B‧‧‧Internet

c, c1 to cn‧‧‧ mobile network signaling data

D1, D2‧‧‧ curves

S11 to S14, S21 to S25, S31 to S34 ‧‧‧ steps

FIG. 1 is a schematic diagram of an artificial intelligence vehicle flow estimation system using mobile network signaling data in the present invention; FIG. 2 is a schematic flowchart of an artificial intelligence vehicle flow estimation method using mobile network signaling data in the present invention; FIG. 3A is a schematic flow chart of an artificial intelligence traffic flow estimation method using mobile network signaling data in the training mode of the present invention; FIG. 3B is an artificial intelligence traffic flow estimation model (such as a neural network such as an LSTM architecture) in the present invention Road model); Figure 4 is a flow diagram of the artificial intelligence vehicle flow estimation method using mobile network signaling data in the application mode of the present invention; and Figure 5 is a flowchart of mobile network signaling data using the present invention Compared with the prior art systems and methods using EVP (Electric Road Pricing System-based Vehicle Detection), the artificial intelligence vehicle flow estimation system and method is a comparison curve diagram for estimating or detecting vehicle flow.

The following describes the embodiments of the present invention with specific specific implementation forms. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this description, and can also be implemented by other different specific implementation forms. Or apply.

The present invention discloses an artificial intelligence traffic flow estimation system and method using mobile network signaling data, which uses unsupervised learning algorithms to extract characteristics of mobile network signaling data, and then uses supervised learning algorithms to train artificial intelligence traffic flows. Estimation models trained in obtaining mobile network signaling data After the artificial intelligence traffic flow estimation model, the trained artificial intelligence traffic flow estimation model can be used to estimate the traffic flow information (such as traffic flow) on the designated road.

The invention provides a system and method for designing a traffic flow estimation system using a plurality of mobile network signaling data (such as CVP signaling data) and artificial intelligence in combination between a plurality of mobile devices and a network. The core technology is an unsupervised algorithm first. The mobile network signaling data is divided into different ethnic groups, and training is performed using a supervised algorithm to generate a trained artificial intelligence traffic flow estimation model, and then the artificial intelligence traffic flow estimation model is used to estimate the traffic flow information of multiple vehicles ( Such as traffic). Therefore, compared with the previous method of directly using the statistics of mobile network signaling data as the traffic flow, the present invention can improve the accuracy of traffic flow information and the inaccuracy of traffic flow information generated by telecommunications operators due to insufficient mobile network occupancy. problem.

FIG. 1 is a schematic diagram of an artificial intelligence traffic flow estimation system 1 using mobile network signaling data c in the present invention. As shown in the figure, the artificial intelligence vehicle flow estimation system 1 includes a mobile network signaling data extraction module 10, an unsupervised learning module 20, a feature extraction module 30, and a supervised learning (Supervised Learning) module 40 and an artificial intelligence traffic flow estimation model 50.

The mobile network signaling data acquisition module 10 can capture a plurality of mobile network signaling data e between a plurality of mobile devices A and at least one network B passing through the designated road within a predetermined range of the designated road. For example, the mobile network signaling data c includes data such as the location, time, and speed of the mobile devices A. The mobile device A may be a smart phone, a smart watch, a SIM (Subscriber Identity Module) card, Tablet, etc., and the mobile network signaling data c may be CVP (vehicle detection based on mobile device base station) signaling data.

The unsupervised learning module 20 may have one of the unsupervised learning algorithms 21 of artificial intelligence to use the unsupervised learning algorithms 21 to extract mobile network signaling data according to the speed of the mobile devices A The captured mobile network signaling data c is divided into a plurality of different ethnic groups.

The feature extraction module 30 may extract at least one (such as three) specific ethnic groups from the different ethnic groups divided by the unsupervised learning algorithm 21 to calculate the unit time of each of the at least one (such as three) specific ethnic groups. The quantity of the mobile network signaling data c is taken as the extracted feature. For example, each time unit may be every minute, every five minutes, every ten minutes, or every fifteen minutes, and so on.

The supervised learning module 40 may have one of the supervised learning algorithms 41 of artificial intelligence to use the supervised learning algorithm 41 to establish an artificial intelligence traffic flow estimation model 50 based on the features extracted by the feature extraction module 30. For example, the supervised learning algorithm 41 may be a Linear Regression algorithm, a Support Vector Machine (SVM) algorithm, a Decision Tree algorithm, a Rcndom Forecast algorithm, Neural-like network, etc., and the neural-like network can be Back-Propagation Neural Network (BPNN), Recurrent Neural Network (RNN), Deep Neural Network (DNN) ), Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM) models, etc.

At the same time, the supervised learning module 40 can extract the feature extraction module 30 The latter feature and the traffic volume of the plural vehicles (not shown) are used as tags to form a training set, so that the training set is repeatedly trained using the supervised learning algorithm 41, thereby reducing the error of mixing these vehicles with plural persons. The traffic flow of the aforementioned vehicles can be determined by the image recognition software of external roadside equipment (such as monitors), the vehicle detector (VD) of roadside equipment (such as radar), and the EVP of electronic tag (e-Tag) equipment (using electronic Road toll system-based vehicle detection), or personnel intersection surveys, etc., provide preliminary source information.

The artificial intelligence traffic flow estimation model 50 may be generated by the supervised learning module 40, and the feature extraction module 30 may input the extracted features into the artificial intelligence traffic flow estimation model 50 to be estimated by the artificial intelligence traffic flow estimation model 50. Estimating or generating traffic flow information for a plurality of vehicles passing a designated road 60.

FIG. 2 is a schematic flowchart of an artificial intelligence vehicle flow estimation method using mobile network signaling data c in the present invention, and please refer to FIG. 1 above. At the same time, the main technical contents of the artificial intelligence vehicle flow estimation method using the mobile network signaling data c in the present invention are as follows, and the remaining technical contents are the same as the detailed description of FIG. 1 described above, and are not repeated here.

In step S11 of FIG. 2, a mobile network signaling data acquisition module 10 retrieves the information between a plurality of mobile devices A and at least one network B that pass through a specified road within a predetermined range of a specified road. Plural mobile network signaling data c.

In step S12 of FIG. 2, an unsupervised learning algorithm 21 with artificial intelligence in an unsupervised learning module 20 is used to retrieve mobile network signaling data according to the speed of the mobile devices A. The mobile network signaling data c captured by the module 10 is divided into a plurality of different ethnic groups.

In step S13 of FIG. 2, a feature extraction module 30 extracts at least one specific ethnic group from different ethnic groups divided by the unsupervised learning algorithm 21, and the feature extraction module 30 calculates at least one specific ethnic group. The number of mobile network signaling data c per unit time in the data is taken as the extracted feature.

In step S14 of FIG. 2, a supervised learning algorithm 41 having artificial intelligence in a supervised learning module 40 is used to establish an artificial intelligence traffic flow estimation model 50 based on the features extracted by the feature extraction module 30. The artificial traffic flow estimation model 50 is used to estimate or generate traffic flow information of a plurality of vehicles passing a designated road.

FIG. 3A is a schematic flow chart of the artificial intelligence vehicle flow estimation method using the mobile network signaling data c in the “training mode” in the present invention, and FIG. 3B is an artificial intelligence vehicle flow estimation model (such as the LSTM architecture in the present invention). The neural network-like model is also applicable to the artificial intelligence traffic flow estimation system 1 in FIG. 1, which is described below.

In step S21 of FIG. 1 and FIG. 3A, the mobile network signaling data acquisition module 10 retrieves a predetermined range of a specified road, and a plurality of mobile devices A and at least one network B passing the specified road. Between the plural historical mobile network signaling data c. For example, the historical mobile network signaling data c includes historical location, time, and speed data of mobile device A, and the historical mobile network signaling data c refers to its earlier mobile network signaling data c .

For example, the mobile network signaling data extraction module 10 was used to capture funds from Fund 3 Road to Fund 2 Road in Keelung City from December 4, 2017 to December 2017. Multiple mobile network signalling information on May 22c. For example, the mobile network signaling data c includes data such as the location, time, and vehicle speed of the mobile devices A.

In step S22 of FIG. 1 and FIG. 3A, the unsupervised learning algorithm 21 of the unsupervised learning module 20 is used to signal the historical mobile network signaling data according to the speed of the mobile devices A c Generate the number of different ethnic groups. For example, the unsupervised learning algorithm 21 may be a K-means clustering algorithm, a hierarchical clustering algorithm, a DBSCAN (density-based spatial clustering of applications with noise; density-based spatial clustering) algorithm, and the like.

For example, the unsupervised learning algorithm 21 is used to generate the number of different groups of the plural mobile network signaling data c. For example, the mobile network signaling data c are plural mobile network signaling data c1, c2, respectively. c3, c4, c5, ..., cn, and the mobile network signaling data c1, c2, c3, c4, c5, ..., cn respectively correspond to vehicle speeds of 10, 20, 30, 40, 50 , ..., N kilometers per hour (km / hr). Assuming that the mobile network signaling data c1, c2, c3, c4, c5, ..., cn are first divided into two ethnic groups, the unsupervised learning algorithm 21 may randomly take the center points of the two ethnic groups, For example, the center points of the two ethnic groups are mobile network signaling data c2 (vehicle speed is 20 km / h) and mobile network signaling data c5 (vehicle speed is 50 km / h).

At the same time, the unsupervised learning algorithm 21 can calculate the distance (such as Euclidean distance) between the remaining mobile network signaling data and the center points of the two groups (such as mobile network signaling data c2 and c5), and The remaining mobile network signaling data are classified into the nearest group. For example: mobile network signaling data c1 The distances from c3 and c3 to the mobile network signaling data c2 are 10 kilometers per hour, and the distances from the mobile network signaling data c1 and c3 to the mobile network signaling data c5 are 20 and 40, respectively. Kilometers, the mobile network signaling data c1, c2, and c3 can be divided into the same group, and the mobile network signaling data c4 and c5 can be divided into another group.

Then, the unsupervised learning algorithm 21 may recalculate the center point of the new ethnic group, and according to the above method, use the center point of the new ethnic group to transmit all the mobile network signaling data c1, c2, c3, c4, c5, ..., The cn is re-divided into different ethnic groups, and iteratively executed until the ethnic groups no longer change. For example, all the mobile network signaling data c1, c2, c3, c4, c5, ..., cn are divided into five ethnic groups.

In step S23 of FIG. 1 and FIG. 3A, the feature extraction module 30 extracts at least one specific ethnic group from the different ethnic groups as one of the characteristics (including the number of majority ethnic groups or the number of paired ethnic groups, etc.) to calculate The number and time step of the mobile network signaling data c per unit time in at least one specific group are generated, and the extracted features are generated. For example, the features extracted by the feature extraction module 30 include the number and time step of the mobile network signaling data c per unit time in at least one specific ethnic group, the number of majority ethnic groups in different ethnic groups, or the number of paired ethnic groups.

For example, the feature extraction module 30 extracts the number of the middle three ethnic groups from the above five ethnic groups as one of the characteristics, so as to calculate the mobile network signaling data c for each unit time in the middle three ethnic groups. Quantities and time steps produce the characteristics after extraction. For example, the mobile network signaling data c is counted every five minutes, and the day is divided into 288 time points to correspond to The number of mobile network signaling data c at 288 time steps. Each time step includes the total number of ethnic groups and the respective numbers of the three middle ethnic groups.

In step S24 of FIG. 1 and FIG. 3A, the image recognition software of the external roadside equipment (such as a monitor), the vehicle detector (VD) of the roadside equipment (such as a radar), and the e-Tag (electronic tag) The equipment's EVP (electronic vehicle toll detection based on electronic road toll system), or the intersection survey of personnel, etc., obtains the detection data within a predetermined range of a specified road in the historical time as a label. Then, the supervised learning module 40 uses the features extracted by the feature extraction module 30 and the traffic volume of a plurality of vehicles as labels to form a training set, and uses the supervised learning algorithm 41 to repeatedly train the training set, thereby reducing the These vehicles are intermingled with multiple people.

For example, the EVP of the e-Tag (electronic tag) device is captured from Fund 3 Road to Fund 2 Road in Keelung City. From December 4, 2017 to December 22, 2017, the EVP data every five minutes is used as the tag. The supervised learning module 40 composes the extracted features and labels into a training set.

In step S25 of FIG. 1 and FIG. 3A, a supervised learning module 40 is used to establish an artificial intelligence traffic flow estimation model 50, and a supervised learning algorithm 41 is used to perform a training set on the artificial intelligence traffic flow estimation model 50. The training and calibration are adjusted to a reasonable error range, and a trained artificial intelligence traffic flow estimation model 50 is generated.

For example, a supervised learning algorithm 41 can be used to build an artificial intelligence traffic flow estimation model 50. For example, the artificial intelligence traffic flow estimation model 50 can be a neural network model such as the LSTM (Long Short-Term Memory) architecture shown in Figure 3B. . At the same time, features (such as the total number of ethnic groups, (The respective numbers of the three ethnic groups) in the intelligent traffic flow estimation model 50, and input the characteristic data from the current time point (t) to the previous five time points (t-5) in order, and then use the supervised learning algorithm Method 41 uses the above training set to train and adjust the artificial intelligence traffic flow estimation model 50 to generate the artificial intelligence traffic flow estimation model 50 after training and adjustment.

The artificial intelligence vehicle flow estimation model 50 (such as a neural network model such as the LSTM architecture) in FIG. 3B described above may include an input layer 51, an LSTM (long-term short-term memory) layer 52, a hidden layer 53 and an output layer 54, Among them, t represents a time step, t-5 represents the next 5 time steps, x1 to x4 represent different features, j1 to jn represent hidden neurons, and y (t) represents the traffic flow.

FIG. 4 is a flow chart of the artificial intelligence traffic flow estimation method using the mobile network signaling data c in the “application mode” in the present invention, and is also applicable to the artificial intelligence traffic flow estimation system 1 of FIG. as follows.

In step S31 of FIG. 1 and FIG. 4, the mobile network signaling data acquisition module 10 retrieves a predetermined range of a specified road, and a plurality of mobile devices A and at least one network B passing the specified road. Plural real-time mobile network signaling data between c. For example, the real-time mobile network signaling data c includes the real-time location, time, and speed of mobile device A.

For example, the mobile network signaling data extraction module 10 captures a plurality of mobile network signaling data from GF 3 to GF 2 in Keelung City on December 23, 2017 from 8:00 to 24:00 c. For example, the mobile network signaling data c includes data such as the location, time, and vehicle speed of the mobile devices A.

In step S32 of FIG. 1 and FIG. 4, the unsupervised learning algorithm 21 of the unsupervised learning module 20 is used, and according to the mobile devices A Speed, the real-time mobile network signaling data c is generated to the number of different ethnic groups. For example, the unsupervised learning algorithm 21 may be a K-means clustering algorithm, a hierarchical clustering algorithm, a DBSCAN (density-based spatial clustering with noise) clustering algorithm, and the like.

For example, the unsupervised learning algorithm 21 (such as the K-means algorithm) of the unsupervised learning module 20 is used to divide the real-time mobile network signaling data c into five groups.

In step S33 of FIG. 1 and FIG. 4, the feature extraction module 30 extracts at least one specific ethnic group from the different ethnic groups as one of the features to calculate an action network per unit time in the at least one specific ethnic group. The number and time step of the channel signalling data c generate the extracted characteristics. For example, the features extracted by the feature extraction module 30 include the number and time step of the mobile network signaling data c per unit time in at least one specific ethnic group, the number of majority ethnic groups in different ethnic groups, or the number of paired ethnic groups.

For example, the feature extraction module 30 extracts the number of the middle three ethnic groups from the above five ethnic groups as one of the characteristics, so as to calculate the mobile network signaling data c for each unit time in the middle three ethnic groups. Quantities and time steps produce the characteristics after extraction. For example, the mobile network signaling data c is counted every five minutes, and the day can be divided into 288 time points to correspond to the number of mobile network signaling data c at 288 time steps. Each time step includes the Total number, each of the three ethnic groups in the middle.

In step S34 of FIG. 1 and FIG. 4, the feature extraction module 30 inputs the extracted features to the trained artificial intelligence vehicle flow estimation model 50 to estimate or Generated by The traffic flow information of the vehicles on the road 60.

Fig. 5 is the "artificial intelligence vehicle flow estimation system and method using mobile network signaling data" of the present invention compared with the prior art "system and method using EVP (vehicle detection based on electronic road pricing system)" A data comparison curve diagram for estimating or detecting vehicle flow, wherein the present invention is a curve D1, and the prior art is a curve D2.

In the prior art systems and methods for detecting vehicle flow using EVP, roadside equipment for detecting e-Tags (electronic tags) needs to be built, and vehicles also need to be equipped with e-Tags, resulting in higher equipment or Hardware construction costs. On the contrary, in the artificial intelligence vehicle flow estimation system and method using mobile network signaling data of the present invention, it is not necessary to build a roadside device for detecting e-Tag (electronic tag), and the vehicle does not need to be installed with e-Tag So as to save or reduce the installation cost of related equipment or hardware. Therefore, the artificial intelligence vehicle flow estimation system and method using mobile network signaling data of the present invention can not only replace the existing systems and methods using EVP to detect vehicle flow, but also can save or reduce the installation cost of related equipment or hardware .

In summary, the artificial intelligence vehicle flow estimation system and method using mobile network signaling data in the present invention may have the following features, advantages, or technical effects:

1. The present invention only needs to use the mobile device of a general user and use the mobile network signaling data to estimate the traffic volume on the road. It is not necessary to use the communication device on the vehicle, and the vehicle does not need to install an e-Tag (electronic tag). There is also no need to build additional roadside equipment (such as a monitor or radar) to detect e-Tags (electronic tags), thereby greatly reducing the cost and time of the installation of related equipment or hardware.

2. Compared with the previous method of directly using the statistics of mobile network signaling data as the traffic flow, the present invention uses artificial intelligence technology to improve the accuracy of traffic flow information and the lack of occupancy rate of telecommunications operators due to the mobile network. The problem is that the traffic information is not accurate.

The above-mentioned embodiments merely exemplify the principles, features, and effects of the present invention, and are not intended to limit the implementable scope of the present invention. Anyone who is familiar with this technology can perform the above operations without departing from the spirit and scope of the present invention. Modifications and changes to the implementation form. Any equivalent changes and modifications made by using the disclosure of the present invention should still be covered by the scope of patent application. Therefore, the scope of protection of the rights of the present invention should be as listed in the scope of patent application.

Claims (20)

An artificial intelligence vehicle flow estimation system using mobile network signaling data includes: a mobile network signaling data acquisition module that captures a plurality of mobile devices that pass through a designated road within a predetermined range of the designated road Plural mobile network signaling data between at least one network; and an unsupervised learning module, which is an unsupervised learning algorithm with artificial intelligence to use the unsupervised learning algorithm based on the mobile devices Speed of the mobile network signaling data extraction module into a plurality of different ethnic groups; a feature extraction module, based on the unsupervised learning algorithm based on the actions The speed of the device extracts at least one specific group from the different groups into which the mobile network signaling data is divided, and calculates the amount of mobile network signaling data per unit time in the at least one specific group as The extracted features; and a supervised learning module, which is one of the supervised learning algorithms of artificial intelligence, to utilize the supervised learning algorithm, which will be determined by the feature After taking the feature extraction module establishes a conjecture AI traffic model, or a way of estimating traffic generated by the vehicle of the plurality of specified information through the road traffic AI conjecture model. According to the artificial intelligence vehicle flow estimation system described in item 1 of the scope of patent application, wherein the mobile network signaling data is CVP (vehicle detection based on mobile device base station) signaling data, and the extracted features include The number and time step of mobile network signaling data per unit time in the at least one specific ethnic group, the number of majority ethnic groups in the different ethnic groups, or the number of paired ethnic groups. The artificial intelligence vehicle flow estimation system as described in the first item of the patent application scope, wherein the unsupervised learning algorithm is a K-means clustering algorithm, a hierarchical clustering algorithm, or a DBSCAN (density-based spatial clustering with noise) ) Algorithm, the supervised learning algorithm is a linear regression algorithm, a support vector machine (SVM) algorithm, a decision tree algorithm, a random forest algorithm, or a neural network-like algorithm. According to the artificial intelligence traffic flow estimation system described in item 1 of the scope of patent application, wherein the supervised learning module further uses the features extracted by the feature extraction module and the traffic volume of the vehicles as labels to form a training set, The training set is repeatedly trained using the supervised learning algorithm. For example, the artificial intelligence traffic flow estimation system described in item 4 of the scope of patent application, wherein the supervised learning algorithm further adjusts the artificial intelligence traffic flow estimation model to a reasonable error range based on the training set, thereby generating the This artificial intelligence traffic flow estimation model is trained. According to the artificial intelligence vehicle flow estimation system described in item 1 of the scope of patent application, in the training mode of the artificial intelligence vehicle flow estimation system, the mobile network signaling data extraction module further captures the designated road Within a predetermined range, the mobile network signaling data of the plural history between the mobile devices on the designated road and the network is utilized to utilize the unsupervised learning algorithm to convert the mobile devices based on the speed of the mobile devices. Historical mobile network signaling data generated the numbers of these different ethnic groups. According to the artificial intelligence traffic flow estimation system described in item 6 of the patent application scope, wherein the feature extraction module further extracts the at least one specific ethnic group from the different ethnic groups as one of the features to calculate the at least one specific ethnic group. The quantity and time step of the mobile network signaling data per unit time in the group do not produce the extracted characteristics. For example, the artificial intelligence traffic flow estimation system described in item 7 of the scope of patent application, wherein the supervised learning module further uses the features extracted by the feature extraction module and the traffic volume of the vehicles as labels to form a training set. Using the supervised learning algorithm to repeatedly train the training set, and using the supervised learning algorithm to train and adjust the artificial intelligence traffic flow estimation model with the training set to a reasonable error range, and then produce the This artificial intelligence traffic flow estimation model is trained. According to the artificial intelligence vehicle flow estimation system described in item 1 of the patent application scope, in the application mode of the artificial intelligence vehicle flow estimation system, the mobile network signaling data extraction module further captures the designated road Within a predetermined range, through the plurality of real-time mobile network signaling data between the mobile devices on the designated road and the network, to utilize the unsupervised learning algorithm based on the speed of the mobile devices, The real-time mobile network signaling data generates the numbers of these different ethnic groups. The artificial intelligence traffic flow estimation system described in item 9 of the scope of patent application, wherein the feature extraction module further extracts the at least one specific ethnic group from the different ethnic groups as one of the features, and calculates the at least one specific ethnic group. The number and time step of mobile network signaling data per unit time in the group to generate extracted features, and the feature extraction module inputs the extracted features to the trained artificial intelligence traffic flow estimation model To estimate or generate traffic flow information of the vehicles passing the designated road through the artificial intelligence traffic flow estimation model. An artificial intelligence vehicle flow estimation method using mobile network signaling data includes: capturing a plurality of mobile network signals between a plurality of mobile devices passing through the designated road and at least one network within a predetermined range of the designated road; Order data; use an unsupervised learning algorithm with artificial intelligence to divide the mobile network signaling data into a plurality of different ethnic groups based on the speed of the mobile devices; from the unsupervised learning algorithm based on the mobile devices The speed extracts at least one specific group from the different groups into which the mobile network signaling data is divided, and calculates the amount of mobile network signaling data per unit time in the at least one specific group as the extracted And using a supervised learning algorithm with artificial intelligence to establish an artificial intelligence traffic flow estimation model based on the extracted features, so as to estimate or generate the traffic passing through the designated road through the artificial intelligence traffic flow estimation model. Traffic information for multiple vehicles. For example, the artificial intelligence vehicle flow estimation method described in item 11 of the scope of patent application, wherein the mobile network signaling data is CVP (vehicle detection based on mobile device base station) signaling data, and the extracted features include The number and time step of mobile network signaling data per unit time in the at least one specific ethnic group, the number of majority ethnic groups in the different ethnic groups, or the number of paired ethnic groups. The artificial intelligence traffic flow estimation method described in the scope of the patent application No. 11, wherein the unsupervised learning algorithm is a K-means clustering algorithm, a hierarchical clustering algorithm, or a DBSCAN (density-based spatial clustering with noise) ) Algorithm, the supervised learning algorithm is a linear regression algorithm, a support vector machine (SVM) algorithm, a decision tree algorithm, a random forest algorithm, or a neural network-like algorithm. The artificial intelligence traffic flow estimation method described in item 11 of the scope of the patent application, further includes using the extracted features and the traffic flow of the vehicles as labels to form a training set to use the supervised learning algorithm for the training. Set for repeated training. The artificial intelligence traffic flow estimation method described in item 14 of the scope of the patent application, further includes using the supervised learning algorithm to adjust the artificial intelligence traffic flow estimation model to a reasonable error range with the training set, thereby generating the This artificial intelligence traffic flow estimation model is trained. According to the artificial intelligence vehicle flow estimation method described in item 11 of the scope of patent application, the method further includes, in the training mode of the artificial intelligence vehicle flow estimation method, capturing within a predetermined range of the designated road, passing the designated road through the designated road. Plural historical mobile network signaling data between the mobile device and the network, using the unsupervised learning algorithm to generate the differences from the historical mobile network signaling data according to the speed of the mobile devices Number of ethnic groups. The artificial intelligence traffic flow estimation method as described in item 16 of the scope of patent application, further comprising extracting the at least one specific ethnic group from the different ethnic groups as one of the features to calculate each unit time in the at least one specific ethnic group. The amount and time step of the mobile network signaling data does not produce the extracted characteristics. The artificial intelligence traffic flow estimation method described in item 17 of the scope of the patent application, further includes using the extracted features and the traffic volume of the vehicles as labels to form a training set to use the supervised learning algorithm to train the training set. The training set is repeatedly trained, and the supervised learning algorithm is used to train and adjust the artificial intelligence traffic flow estimation model with the training set to a reasonable error range, thereby generating the trained artificial intelligence traffic flow estimation model. According to the artificial intelligence vehicle flow estimation method described in item 11 of the scope of patent application, the method further includes, in the application mode of the artificial intelligence vehicle flow estimation method, capturing within a predetermined range of the designated road, passing the designated road through the designated road. Plural real-time mobile network signaling data between the mobile device and the network, using the unsupervised learning algorithm to generate the differences from the real-time mobile network signaling data according to the speed of the mobile devices Number of ethnic groups. The artificial intelligence traffic flow estimation method described in item 19 of the scope of patent application, further includes extracting the at least one specific ethnic group from the different ethnic groups as one of the features, and calculating each unit time in the at least one specific ethnic group. The amount and time step of the mobile network signaling data to generate extracted features, and then input the extracted features to the trained artificial intelligence traffic flow estimation model to estimate through the artificial intelligence traffic flow estimation model Traffic information of the vehicles passing through the designated road is generated or generated.