Explanatable road network selection method and device, medium and equipment

Abstract

The invention discloses an interpretable road network selection method and device, a storage medium and electronic equipment. The method comprises the following steps: acquiring description features of each test road, and constructing feature vectors based on the description features; determining a teacher model, and sorting the feature importance of the description features input into the teacher model through an interpretable tool; constructing an additive neural network; calculating a knowledge distillation loss function based on a prediction result of the teacher model; the ranked description features serve as a training set to be input into an additive neural network, the additive neural network is trained through a knowledge distillation loss function, the trained additive neural network serves as an explainable road network selection model, and road selection is conducted through the explainable road network selection model. According to the method, high accuracy of the road selection model can be considered, and the road selection model has interpretability.

Description

Technical Field

[0001] The present application relates to the field of geographic information processing technology, and in particular to an interpretable road network selection method, device, storage medium and electronic device. Background Art

[0002] Cartographic generalization refers to the process of deriving map data that expresses details at a large scale from map data that expresses an overview at a smaller scale, thereby constructing map products of different scales, such as topographic maps of different scales and network navigation maps at different levels. As a basic geographical element, the road network is one of the main objects of cartographic generalization. Cartographic generalization of road network data is mainly reflected in the selection of road targets, that is, selecting some important road targets from the original road data to meet the clarity requirements of smaller-scale expression. Among them, the number of road targets selected can be calculated using the square root model. However, the specific road targets to be selected are determined by the importance of each road. It is necessary to comprehensively consider factors such as road attribute level, topological connectivity, geometric length, etc. It is a complex multi-feature analysis and decision-making problem.

[0003] Under the existing technological framework, the road network selection process still requires the intensive involvement of professional cartographers, requiring significant investment and a long lead time, severely hindering the iterative updates of map products. In recent years, researchers have introduced machine learning models to develop intelligent road network selection methods to alleviate or even replace manual involvement in the comprehensive task of road network mapping. These methods typically take various descriptive features of road targets as input, construct a classifier using a machine learning model, and output a selection decision. The classifier's decision-making knowledge is primarily acquired through supervised learning of sample case data. In practical applications, while this machine learning-based selection method can achieve high accuracy, the models employed are often complex, resembling "black boxes," resulting in poor interpretability and difficulty understanding the model's internal decision logic and rules. This poor model interpretability can further hinder the ability to improve model performance as needed, leading to poor application results and low user trust. Summary of the Invention

[0004] The embodiments of the present application provide an explainable road network selection method, device, storage medium and electronic device, which can improve the accuracy of road network selection and enhance user trust due to the explainability of the road selection model.

[0005] The present application provides an interpretable road network selection method, including: Obtaining descriptive features of each test road and constructing a feature vector based on the descriptive features; Determining a teacher model and ranking the feature importance of the descriptive features input into the teacher model by an interpretable tool; Constructing additive neural networks; Calculating a knowledge distillation loss function based on the prediction results of the teacher model; The sorted description features are input into the additive neural network as a training set, the additive neural network is trained using the knowledge distillation loss function, the trained additive neural network is used as an interpretable road network selection model, and road selection is performed using the interpretable road network selection model.

[0006] Furthermore, in the above-mentioned interpretable road network selection method, the step of determining the teacher model includes: The feature vector is used as a training set and a test set, and multiple different machine learning models are trained using the training set. The test set is input into the trained multiple different machine learning models to obtain prediction results, and the machine learning model with the highest prediction result accuracy is selected as the teacher model.

[0007] Furthermore, in the above-mentioned interpretable road network selection method, the step of ranking the feature importance of the descriptive features input into the teacher model by an interpretable tool comprises: Calculate the j The average contribution of a feature to all test samples:

[0008] in, Indicates the total number of test roads, Indicates the i The first target test road j SHAP contribution value of each feature; The description features are sorted in descending order according to the SHAP contribution values.

[0009] Furthermore, in the above-mentioned interpretable road network selection method, the additive neural network includes an input layer, an intermediate layer, and an output layer, and the output of the additive neural network is used to determine whether the test road is selected and retained; The mathematical expression of the additive neural network is:

[0010] in, Represents the feature neural network, n Indicates the number of feature networks, and its maximum value is the total number of described features.

[0011] Furthermore, in the above-mentioned interpretable road network selection method, the step of calculating the knowledge distillation loss function based on the prediction results of the teacher model includes: Predicting a first probability of being selected and retained for each of the test roads by the teacher model, and marking the first probability as a first label; Calculating the soft label of the test road required for knowledge distillation based on the first label is selected and retained; A knowledge distillation loss function is calculated based on the soft labels.

[0012] Furthermore, the above-mentioned interpretable road network selection method further comprises: The soft tag is calculated using a first formula, which is:

[0013] in, is the first label, is the distillation temperature.

[0014] The knowledge distillation loss function is calculated by the second formula, which is:

[0015] in, is the prediction result of the additive neural network, To manually label the test roads, is the weight parameter of soft label loss and true label loss, is the mean square error function.

[0016] Furthermore, in the above-mentioned interpretable road network selection method, the descriptive features include road description features and road connectivity features, the road description features include at least road grade, number of lanes, road surface material, length, width, and number of arcs, and the road connectivity features include at least degree centrality, clustering coefficient, betweenness centrality, closeness centrality, and eigenvector centrality.

[0017] The present application also provides an interpretable road network selection device, including: an acquisition module, configured to acquire descriptive features of each test road and construct a feature vector based on the descriptive features; a determination and ranking module, configured to determine a teacher model and rank the feature importance of the descriptive features input into the teacher model using an interpretable tool; Building blocks for network construction of additive neural networks; A calculation module, configured to calculate a knowledge distillation loss function based on the prediction results of the teacher model; The road selection module is used to input the sorted description features as a training set into the additive neural network, train the additive neural network through the knowledge distillation loss function, use the trained additive neural network as an interpretable road network selection model, and perform road selection through the interpretable road network selection model.

[0018] An embodiment of the present application also provides a computer-readable storage medium, in which a plurality of instructions are stored. The instructions are suitable for being loaded by a processor to execute any of the above-mentioned interpretable road network selection methods.

[0019] An embodiment of the present application also provides an electronic device, including a processor and a memory, wherein the processor is electrically connected to the memory, the memory is used to store instructions and data, and the processor is used to perform the steps in any of the above-mentioned interpretable road network selection methods.

[0020] This application provides an interpretable road network selection method, device, storage medium, and electronic device. This application constructs a teacher model and trains a student model (an additive neural network) based on the teacher model's predictions and a knowledge distillation loss function, resulting in an interpretable road selection model. The additive neural network inherently provides interpretability between input features and output results. Furthermore, knowledge distillation is used to impart the highly accurate teacher model's decision-making knowledge to the additive neural network, enabling the road selection model to achieve both high accuracy and interpretability. Using this road selection model for road selection improves the accuracy of road network selection, and the interpretability of the road selection model enhances user trust. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The following detailed description of the specific embodiments of the present application in conjunction with the accompanying drawings will make the technical solutions and other beneficial effects of the present application apparent.

[0022] Figure 1 A flowchart of an explainable road network selection method provided in an embodiment of the present application.

[0023] Figure 2 A schematic diagram of the feature importance ranking results of the description features provided in the embodiments of the present application.

[0024] Figure 3 A schematic diagram of the structure of the additive neural network provided in an embodiment of the present application.

[0025] Figure 4 A flowchart of a model trained using the knowledge distillation method provided in an embodiment of the present application.

[0026] Figure 5A graph showing the relationship between the feature description and contribution value provided in the embodiment of the present application.

[0027] Figure 6 A schematic diagram of the structure of an interpretable road network selection device provided in an embodiment of the present application.

[0028] Figure 7 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application.

[0029] Figure 8 Another structural schematic diagram of the electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0030] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0031] Embodiments of the present application provide an interpretable road network selection method, device, storage medium, and electronic device. The interpretable road network selection device provided in embodiments of the present application can be integrated into an electronic device, such as a terminal or server. The terminal can include a tablet computer, laptop computer, personal computer (PC), microprocessor, or other device.

[0032] See also Figure 1 , Figure 1 A flowchart of an interpretable road network selection method provided in an embodiment of the present application, which is applied to an electronic device, includes the following steps: S1, obtain the descriptive features of each test road and construct a feature vector based on the descriptive features.

[0033] First, given a regional road network , Indicates the test road, m Indicates the number of test roads included in the road network. It is composed of connected road arcs with the same name.

[0034] Then, get each test road Describe the features and construct the feature vector, recorded as , KRepresents the number of features. The road description features used include two aspects: first, road description features, including road grade, number of lanes, pavement material, length, width, and number of arcs; second, the connectivity characteristics of each road within the entire road network. These can be defined and calculated based on a road network graph model (i.e., road objects are graph nodes, and the connections between road objects are expressed as edges). These characteristics include degree centrality, clustering coefficient, betweenness centrality, closeness centrality, and eigenvector centrality. The eigenvectors are then normalized.

[0035] S2, determines the teacher model and ranks the feature importance of the descriptive features input into the teacher model through interpretable tools.

[0036] In one embodiment, step S2 includes the following steps: S21, use the feature vector as the training set and the test set, train multiple different machine learning models through the training set, and input the test set into the trained multiple different machine learning models to obtain the prediction results, and select the machine learning model with the highest prediction result accuracy as the teacher model.

[0037] Specifically, each road is manually labeled ("keep" or "discard") and divided into training and test sets according to a certain ratio (e.g., 8:2). The training set is fed into different machine learning models to obtain prediction results, and training is then iterated based on the labels. The test set is then fed into the trained machine learning models to obtain prediction results. The machine learning model with the highest prediction accuracy is selected as the teacher model.

[0038] As an example, the machine learning model may include a classifier model such as a random forest, a gradient boosting tree, or a BP neural network.

[0039] S22, calculate the j The average contribution of a feature to all test samples:

[0040] in, Indicates the total number of test roads, Indicates the i The first target test road j SHAP contribution value of each feature.

[0041] Among them, the SHAP contribution value is calculated by the SHAP (SHapley Additive exPlanations) interpretable tool.

[0042] S23, sort the descriptive features in descending order according to the SHAP contribution value.

[0043] Figure 2 The following is a schematic diagram of the feature importance ranking result of the description features provided in the embodiment of the present application, such as Figure 2 As shown in Figure 3, the sorting results are rank, number of arcs, degree centrality, length, betweenness centrality, closeness centrality, eigenvector centrality, width, clustering coefficient, material, and number of lanes. S3, construct an additive neural network.

[0044] Figure 3 The schematic diagram of the structure of the additive neural network provided in the embodiment of the present application is as follows: Figure 3 As shown in the figure, the Neural Additive Model (NAM) consists of multiple feature neural networks, each of which consists of an input layer, an intermediate layer, and an output layer. The input layer is a single neuron that controls the input of road features. The intermediate layer contains a few neurons and uses the Relu activation function to ensure its nonlinear fitting ability. The output layer is a single neuron and uses the Sigmoid activation function. Each feature neural network only accepts a single feature input and outputs the contribution value of that feature. ,and The additive neural network is used to determine whether the test road is selected and retained. n The additive neural network of the feature neural network is denoted as , a test road n Descriptive feature input , output the prediction result of the test road, that is = In actual application, you can set , which means that the road target is selected and retained.

[0045] The mathematical expression of additive neural network is:

[0046] in, Represents the feature neural network, n It represents the number of feature networks, and its maximum value is the total number K of features described.

[0047] S4, calculates the knowledge distillation loss function based on the prediction results of the teacher model.

[0048] Figure 4 The flowchart of the model training method by the knowledge distillation method provided in the embodiment of this application can be referred to Figure 4 , step S4 comprises the following steps: S41, predicting a first probability of being selected and retained for each test road through the teacher model, and marking the first probability as a first label.

[0049] S42: Calculate the soft labels of the test roads required for knowledge distillation based on the first labels and select and retain them.

[0050] Specifically, the sorted descriptive features are input into the trained teacher model to obtain the prediction result of the teacher model, that is, the probability of each test road being selected and retained. The first probability of the teacher model predicting that each test road is selected and retained is obtained and recorded as the first label. , the test road prediction required for knowledge distillation is calculated by the following formula to select the retained soft label , where T is the distillation temperature. The larger the value of T, the The predicted probability distribution of different road targets is smoother.

[0051]

[0052] in, is the first label, is the distillation temperature.

[0053] S43, calculate the knowledge distillation loss function based on the soft labels.

[0054] The knowledge distillation loss is composed of the weighted addition of the hard label (artificial label) loss and the soft label loss, as shown in the following formula:

[0055] in, is the prediction result of the additive neural network, The artificial label for the test road is a hard-coded 0 or 1. is the weight parameter of soft label loss and true label loss, is the mean square error function.

[0056] The setting of the Loss function makes the predicted output of the student model as close to the prediction of the teacher model as possible while approaching the true label, thereby transferring the decision-making knowledge of the teacher model to the student model.

[0057] S5: Input the sorted description features as a training set into the additive neural network, train the additive neural network through the knowledge distillation loss function, use the trained additive neural network as an interpretable road network selection model, and perform road selection through the interpretable road network selection model.

[0058] Specifically, different feature quantity combinations are constructed according to the order of the descriptive features, and different feature quantity combinations are input into the additive neural network model (denoted as ( ), combined with the selected teacher model and loss function The training is performed using the knowledge distillation method, and the best performing model on the test dataset is selected as the final interpretable road network selection model.

[0059] It should be explained that different combinations of feature quantities refer to the combinations of the top n features in terms of importance in feature ranking. For example, combination 1 is [rank], combination 2 is [rank, number of radians], and combination 3 is [rank, number of radians, degree centrality].

[0060] The road network selection model obtained by the above method is interpretable. Figure 5 The relationship between the feature description and contribution value provided in the embodiment of this application is as follows: Figure 5 As shown, the final interpretable road network selection model is , that is, the model utilizes the four most important features in feature sorting: level, number of arcs, degree centrality, and length, and achieves its optimal accuracy. We visualize the four feature network curves: the horizontal axis is the normalized size of the feature value, the vertical axis is the feature contribution, and the blue dots represent the overall road target feature distribution. Then, any feature of a road target can find its explainable contribution value, such as Figure 5 Indicated by the green dot.

[0061] The interpretable road network selection method provided in this application constructs a teacher model and trains a student model (an additive neural network) based on the teacher model's predictions and a knowledge distillation loss function, resulting in an interpretable road selection model. The additive neural network inherently provides interpretability between input features and output results. Furthermore, knowledge distillation is used to impart the highly accurate decision-making knowledge of the teacher model to the additive neural network, enabling the road selection model to achieve both high accuracy and interpretability. Using this road selection model for road selection not only ensures high road network selection accuracy, but also enhances user trust due to its interpretability.

[0062] According to the method described in the above embodiment, this embodiment will be further described from the perspective of an interpretable road network selection device. The interpretable road network selection device can be implemented as an independent entity or integrated into an electronic device. The electronic device can be a terminal, server, etc., wherein the terminal can include a tablet computer, a laptop computer, a personal computer (PC), a micro processing box, or other devices.

[0063] See also Figure 6 , Figure 6 The present invention specifically describes an interpretable road network selection device provided by an embodiment of the present invention, which is applied to an electronic device. The interpretable road network selection device may include: An acquisition module is used to obtain the descriptive features of each test road and construct a feature vector based on the descriptive features; The determination and ranking module is used to determine the teacher model and rank the feature importance of the descriptive features input into the teacher model through interpretable tools; Building blocks for network construction of additive neural networks; The calculation module is used to calculate the knowledge distillation loss function based on the prediction results of the teacher model; The road selection module is used to input the sorted descriptive features as a training set into the additive neural network, train the additive neural network through the knowledge distillation loss function, use the trained additive neural network as an interpretable road network selection model, and perform road selection through the interpretable road network selection model.

[0064] During specific implementation, the above modules and / or units can be implemented as independent entities, or can be arbitrarily combined to be implemented as the same or several entities. The specific implementation of the above modules and / or units can refer to the previous method embodiments. The specific beneficial effects that can be achieved can also be found in the beneficial effects in the previous method embodiments, which will not be repeated here.

[0065] In addition, an embodiment of the present application also provides an electronic device, which may be a computer, a tablet computer, or other device. Figure 7 As shown, the electronic device 400 includes a processor 401 and a memory 402. The processor 401 is electrically connected to the memory 402.

[0066] The processor 401 is the control center of the electronic device 400. It uses various interfaces and lines to connect various parts of the entire electronic device. By running or loading applications stored in the memory 402 and calling data stored in the memory 402, it executes various functions of the electronic device and processes data, thereby monitoring the electronic device as a whole.

[0067] In this embodiment, the processor 401 in the electronic device 400 loads instructions corresponding to one or more application processes into the memory 402 according to the following steps, and the processor 401 runs the application stored in the memory 402 to implement various functions: Obtaining descriptive features of each test road and constructing a feature vector based on the descriptive features; Determining a teacher model and ranking the feature importance of the descriptive features input into the teacher model by an interpretable tool; Constructing additive neural networks; Calculating a knowledge distillation loss function based on the prediction results of the teacher model; The sorted description features are input into the additive neural network as a training set, the additive neural network is trained using the knowledge distillation loss function, the trained additive neural network is used as an interpretable road network selection model, and road selection is performed using the interpretable road network selection model.

[0068] The electronic device can implement the steps in any embodiment of the interpretable road network selection method provided in the embodiments of the present application. Therefore, it can achieve the beneficial effects that can be achieved by any interpretable road network selection method provided in the embodiments of the present invention. Please refer to the previous embodiments for details and will not be repeated here.

[0069] Figure 8 The following is a block diagram of the specific structure of an electronic device provided in an embodiment of the present invention. This electronic device can be used to implement the interpretable road network selection method provided in the above embodiments. The electronic device 500 can be a terminal, server, or other device. The terminal can include a tablet computer, laptop computer, personal computer (PC), microprocessor, or other device.

[0070] RF circuit 510 is used to receive and transmit electromagnetic waves, converting them into electrical signals, thereby enabling communication with a communications network or other devices. RF circuit 510 may include various existing circuit components for performing these functions, such as an antenna, a radio frequency transceiver, a digital signal processor, an encryption / decryption chip, a subscriber identity module (SIM) card, memory, and the like. RF circuit 510 can communicate with various networks, such as the Internet, an intranet, or a wireless network, or with other devices via a wireless network. These wireless networks may include cellular telephone networks, wireless local area networks, or metropolitan area networks. The wireless networks may utilize various communication standards, protocols, and technologies, including but not limited to Global System for Mobile Communication (GSM), Enhanced Data GSM Environment (EDGE), Wideband Code Division Multiple Access (WCDMA), Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Wireless Fidelity (Wi-Fi) (such as Institute of Electrical and Electronics Engineers standards IEEE 802.11a, IEEE 802.11b, IEEE802.11g, and / or IEEE802.11n), Voice over Internet Protocol (VoIP), Worldwide Interoperability for Microwave Access (Wi-Max), other protocols for email, instant messaging, and short messaging, and any other suitable communication protocols, including those currently undeveloped.

[0071] The memory 520 can be used to store software programs and modules, such as the corresponding program instructions / modules in the above-mentioned embodiments. The processor 580 executes various functional applications and data processing by running the software programs and modules stored in the memory 520, that is, realizing functions such as taking pictures with the front camera, processing the captured images, and switching the display color of the displayed content on the display screen. The memory 520 may include a high-speed random access memory and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 520 may further include a memory remotely located relative to the processor 580, and these remote memories may be connected to the electronic device 500 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0072] The input unit 530 may be used to receive input digital or character information, and generate a keyboard and a mouse related to user settings and function control. The display unit 540 can be used to display information input by the user or information provided to the user, as well as various graphical user interfaces. These graphical user interfaces can be composed of graphics, text, icons, videos, or any combination thereof. The display unit 540 may include a display panel 541. Optionally, the display panel 541 can be configured in the form of an LCD (Liquid Crystal Display), an OLED (Organic Light-Emitting Diode), or the like.

[0073] Audio circuit 560, speaker 561, and microphone 562 provide an audio interface between the user and electronic device 500. Audio circuit 560 converts received audio data into electrical signals and transmits them to speaker 561, which then converts them into sound signals for output. Microphone 562, on the other hand, converts collected sound signals into electrical signals, which are then received by audio circuit 560 and converted into audio data. The audio data is then processed by output processor 580 and transmitted via RF circuit 510 to, for example, another terminal. Alternatively, the audio data may be output to memory 520 for further processing. Audio circuit 560 may also include an earphone jack to allow communication between external headphones and electronic device 500.

[0074] Electronic device 500, through a transmission module 570 (e.g., a Wi-Fi module), can help users receive requests, send information, and so on, providing users with wireless broadband Internet access. Although the figure shows transmission module 570, it is understood that it is not a required component of electronic device 500 and can be omitted as needed without changing the essence of the invention.

[0075] Processor 580 is the control center of electronic device 500. It connects all components of the phone using various interfaces and circuits. By running or executing software programs and / or modules stored in memory 520 and accessing data stored in memory 520, it executes various functions of electronic device 500 and processes data, thereby providing overall monitoring of the electronic device. Optionally, processor 580 may include one or more processing cores. In some embodiments, processor 580 may integrate an application processor and a modem processor. The application processor primarily handles the operating system, user interface, and application programs, while the modem processor primarily handles wireless communications. It is understood that the modem processor may not be integrated into processor 580.

[0076] Electronic device 500 also includes a power supply 590 (e.g., a battery) for powering various components. In some embodiments, the power supply can be logically connected to processor 580 via a power management system, thereby enabling the power management system to manage charging, discharging, and power consumption. Power supply 590 can also include any components, such as one or more DC or AC power supplies, a recharging system, a power failure detection circuit, a power converter or inverter, and a power status indicator.

[0077] Although not shown, the electronic device 500 also includes a camera (such as a front camera and a rear camera), a Bluetooth module, etc., which will not be described in detail here. Specifically, in this embodiment, the display unit of the electronic device is a touch screen display, and the mobile terminal also includes a memory and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by one or more processors. The one or more programs include instructions for performing the following operations: Obtaining descriptive features of each test road and constructing a feature vector based on the descriptive features; Determining a teacher model and ranking the feature importance of the descriptive features input into the teacher model by an interpretable tool; Constructing additive neural networks; Calculating a knowledge distillation loss function based on the prediction results of the teacher model; The sorted description features are input into the additive neural network as a training set, the additive neural network is trained using the knowledge distillation loss function, the trained additive neural network is used as an interpretable road network selection model, and road selection is performed using the interpretable road network selection model.

[0078] During specific implementation, the above modules can be implemented as independent entities, or can be arbitrarily combined and implemented as the same or several entities. The specific implementation of the above modules can be found in the previous method embodiments and will not be repeated here.

[0079] Those skilled in the art will appreciate that all or part of the steps in the various methods of the above embodiments can be accomplished through instructions, or by controlling related hardware through instructions. The instructions can be stored in a computer-readable storage medium and loaded and executed by a processor. To this end, an embodiment of the present invention provides a storage medium storing a plurality of instructions that can be loaded by a processor to execute the steps of any of the embodiments of the interpretable road network selection method provided in the embodiments of the present invention.

[0080] The computer-readable storage medium may include a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0081] Since the instructions stored in the storage medium can execute the steps in any embodiment of the interpretable road network selection method provided in the embodiments of the present invention, the beneficial effects that can be achieved by any interpretable road network selection method provided in the embodiments of the present invention can be achieved. Please refer to the previous embodiments for details and will not be repeated here.

[0082] The above is a detailed introduction to an interpretable road network selection method, device, storage medium and electronic device provided in the embodiments of the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for technical personnel in this field, based on the ideas of the present application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. An interpretable road network selection method, characterized in that: The method comprises: Obtaining descriptive features of each test road, and constructing a feature vector based on the descriptive features; determining a teacher model and ranking the feature importance of the descriptive features input into the teacher model by an interpretable tool; Constructing additive neural networks; Calculating a knowledge distillation loss function based on the prediction results of the teacher model; The sorted description features are input into the additive neural network as a training set, the additive neural network is trained through the knowledge distillation loss function, the trained additive neural network is used as an interpretable road network selection model, and road selection is performed through the interpretable road network selection model.

2. The interpretable road network selection method according to claim 1, characterized in that: The determining of the teacher model comprises: The feature vector is used as a training set and a test set, a plurality of different machine learning models are trained through the training set, and the test set is input into the trained plurality of different machine learning models to obtain prediction results, and the machine learning model with the highest prediction result accuracy is selected as the teacher model.

3. The interpretable road network selection method according to claim 1, characterized in that: The step of ranking the feature importance of the descriptive features input into the teacher model by an interpretable tool comprises: Calculate the j The average contribution of a feature to all test samples: in, represents the total number of test roads, Indicates i The first target test road j SHAP contribution value of each feature; The description features are sorted in descending order according to the SHAP contribution values.

4. The interpretable road network selection method according to claim 1, characterized in that: The additive neural network includes an input layer, an intermediate layer and an output layer, and the output of the additive neural network is used to determine whether the test road is selected and retained; The mathematical expression of the additive neural network is: in, represents the feature neural network, n Represents the number of feature networks, and its maximum value is the total number of described features.

5. The interpretable road network selection method according to claim 1, characterized in that: The calculating of the knowledge distillation loss function based on the prediction result of the teacher model includes: Predicting a first probability of being selected and retained for each of the test roads by the teacher model, and marking the first probability as a first label; The soft label of the test road selected and retained that is required for calculating knowledge distillation based on the first label; A knowledge distillation loss function is calculated based on the soft labels.

6. The interpretable road network selection method according to claim 5, characterized in that: The method further comprises: The soft tag is calculated by a first formula, wherein the first formula is: in, is the first label, is the distillation temperature. The knowledge distillation loss function is calculated by the second formula, and the second formula is: in, is the prediction result of the additive neural network, To test the manual labeling of roads, is the weight parameter of soft label loss and true label loss, is the mean square error function.

7. The interpretable road network selection method according to claim 1, characterized in that: The descriptive features include road description features and road connectivity features. The road description features include at least road grade, number of lanes, road surface material, length, width, and number of arcs. The road connectivity features include at least degree centrality, clustering coefficient, betweenness centrality, closeness centrality, and eigenvector centrality.

8. An interpretable road network selection device, characterized in that: include: An acquisition module, used for acquiring descriptive features of each test road and constructing a feature vector based on the descriptive features; A determination and ranking module, for determining a teacher model and ranking the feature importance of the descriptive features input into the teacher model through an interpretable tool; Building blocks for network construction of additive neural networks; A calculation module, used for calculating a knowledge distillation loss function based on the prediction result of the teacher model; A road selection module is used to input the sorted description features as a training set into the additive neural network, train the additive neural network through the knowledge distillation loss function, use the trained additive neural network as an interpretable road network selection model, and perform road selection through the interpretable road network selection model.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor to execute the interpretable road network selection method according to any one of claims 1 to 7.

10. An electronic device, characterized in that: It comprises a processor and a memory, wherein the processor is electrically connected to the memory, the memory is used to store instructions and data, and the processor is used to execute the steps in the interpretable road network selection method described in any one of claims 1 to 7.

Images