Parking space recommendation method, device, electronic device and readable storage medium

Abstract

The present application discloses a parking space recommendation method, device, electronic device and readable storage medium, and relates to the field of neural network technology. Residual feature extraction is performed on the vector to be extracted through multiple residual blocks. The residual blocks can effectively alleviate the problems of gradient vanishing and gradient explosion, and the multiple residual blocks are connected in sequence, which can increase the depth of the parking space recommendation model, so as to help the parking space recommendation model better utilize context information, thereby improving the expression ability and performance of the parking space recommendation model. Then, the residual feature vector is updated through the long short-term memory module, and the dependency relationship between each information in the parking demand information can be captured through the long short-term memory module, so as to further improve the performance of the parking space recommendation model, improve the prediction ability of the parking space recommendation model, and help the parking space recommendation model predict the target parking space that meets the user's needs. It realizes automatic recommendation of parking spaces that meet the user's needs to the user based on the parking demand information.

Description

Technical Field

[0001] The present application relates to the field of neural network technology, and in particular to a parking space recommendation method, device, electronic device and readable storage medium. Background Art

[0002] With the development of technology, cars have become an important tool for most families, and with it comes the problem of limited parking resources. To address this problem, many travel service providers will cooperate with parking lot operators to collect parking space information resources from various parking lots, and push information to nearby parking lots based on the user's travel location, so that users can make travel plans and reduce travel anxiety.

[0003] However, the current parking space recommendation method only recommends vacant parking spaces close to the user's location to the user, which cannot meet the personalized parking needs of different groups of people. Summary of the invention

[0004] The present application aims to solve at least one of the technical problems existing in the prior art. To this end, the present application proposes a parking space recommendation method, device, electronic device and readable storage medium, which can automatically recommend parking spaces that meet the user's needs based on parking demand information.

[0005] According to the first aspect of the present application, the parking space recommendation method includes:

[0006] In response to receiving a parking space search request instruction sent by a user terminal, determining parking demand information based on the parking space search request instruction, the parking demand information including user information, current time information, user target location information, user current location information, intended parking duration, multiple current candidate parking spaces, and parking space price information of the current candidate parking spaces;

[0007] Constructing a vector to be extracted based on the user information, the current time information, the planned parking duration, the parking space price information, the user target location information and the user current location information;

[0008] Inputting the vector to be extracted into a trained parking space recommendation model; the parking space recommendation model includes a long short-term memory module, a fully connected layer, a classifier, and a plurality of residual blocks connected in sequence;

[0009] Performing residual feature extraction on the vector to be extracted by using a plurality of the residual blocks to obtain a residual feature vector;

[0010] Inputting the residual feature vector into the long short-term memory module to obtain an updated feature vector;

[0011] By means of the fully connected layer, the updated feature vector is associated with each of the current candidate parking spaces to obtain a plurality of associated feature vectors;

[0012] Inputting a plurality of the associated feature vectors into the classifier to obtain a probability distribution of the associated feature vectors;

[0013] Based on the probability distribution, taking the current candidate parking space corresponding to the associated feature vector with the largest probability value as the target parking space;

[0014] Send relevant information of the target parking space to the user terminal.

[0015] According to the parking space recommendation method of the embodiment of the present application, at least the following beneficial effects are achieved: the parking space recommendation method of the embodiment of the present application, when receiving the parking space search request instruction sent by the user terminal, determines the parking demand information based on the parking space search request instruction, then constructs a vector to be extracted based on the parking demand information, and inputs the vector to be extracted into the parking space recommendation model. In the parking space recommendation model, residual features are extracted from the vector to be extracted through multiple residual blocks, and the residual blocks can effectively alleviate the problems of gradient disappearance and gradient explosion, and multiple residual blocks are connected in sequence, which can increase the depth of the parking space recommendation model, so as to help the parking space recommendation model better utilize context information, thereby improving the expression ability and performance of the parking space recommendation model. Then, the residual feature vector is updated through the long short-term memory module, and the dependency between each piece of information in the parking demand information can be captured through the long short-term memory module, so as to further improve the performance of the parking space recommendation model, improve the prediction ability of the parking space recommendation model, and help the parking space recommendation model predict the target parking space that meets the user's needs. After obtaining the updated feature vector through the long short-term memory module, the updated feature vector is associated with each current candidate parking space through the fully connected layer to obtain multiple associated feature vectors; the multiple associated feature vectors are input into the classifier to obtain the probability distribution of the associated feature vectors; based on the probability distribution, the current candidate parking space corresponding to the associated feature vector with the largest probability value is used as the target parking space, and the relevant information of the target parking space is sent to the user terminal. In this way, the present application realizes the automatic recommendation of parking spaces that meet the user's needs based on the parking demand information.

[0016] According to some embodiments of the first aspect of the present application, the residual block includes a first extended causal convolution layer, a channel attention module, a spatial attention module, a multiplier, a second extended causal convolution layer, and a residual sum layer;

[0017] The input vector of the residual block located at the head among the residual blocks connected in sequence is the vector to be extracted;

[0018] The output vector of the residual block located at the end of the residual blocks connected in sequence is used as the residual feature vector;

[0019] The processing process of each residual block is:

[0020] Inputting the input vector of the residual block to the first extended causal convolution layer to obtain a first convolution vector;

[0021] Inputting the first convolution vector into the channel attention module to obtain a channel attention feature vector;

[0022] Inputting the channel attention feature vector into the spatial attention module to obtain a spatial attention feature vector;

[0023] Inputting the first convolution vector and the spatial attention feature vector into the multiplier to obtain a multiplication vector;

[0024] Inputting the multiplication vector into the second extended causal convolutional layer to obtain a second convolution vector;

[0025] Through the residual sum layer, residual sum processing is performed on the input vector of the residual block and the second convolution vector to obtain an output vector.

[0026] According to some embodiments of the first aspect of the present application, constructing a vector to be extracted based on the user information, the current time information, the planned parking duration, the parking space price information, the user target location information, and the user current location information includes:

[0027] Converting the unique code in the user information into a first array;

[0028] Converting the coordinate values ​​in the user target location information into a second array;

[0029] Converting the coordinate values ​​in the user's current location information into a third array;

[0030] Converting the parking space price information into a fourth array;

[0031] Converting the current time information into a fifth array;

[0032] Converting the planned parking duration into a sixth array;

[0033] The vector to be extracted is obtained based on the first array, the second array, the third array, the fourth array, the fifth array and the sixth array.

[0034] According to some embodiments of the first aspect of the present application, the residual block further includes a first weight normalization layer and a first rectified linear unit;

[0035] Before inputting the first convolution vector into the channel attention module, comprising:

[0036] Performing a weight normalization operation on the first convolution vector through the first weight normalization layer to obtain a first weight normalized feature vector;

[0037] performing nonlinear activation on the first weight-normalized feature vector by using the first rectified linear unit to obtain a first activated feature vector;

[0038] The step of inputting the first convolution vector into the channel attention module to obtain a channel attention feature vector includes:

[0039] The first activation feature vector is input into the channel attention module to obtain the channel attention feature vector.

[0040] According to some embodiments of the first aspect of the present application, the residual block further includes a second weight normalization layer and a second rectified linear unit;

[0041] Before performing residual sum processing on the input vector of the residual block and the second convolution vector through the residual sum layer, the method further comprises:

[0042] Performing a weight normalization operation on the second convolution vector through the second weight normalization layer to obtain a second weight normalized feature vector;

[0043] performing nonlinear activation on the second weight-normalized feature vector by using the second rectified linear unit to obtain a second activated feature vector;

[0044] The step of performing residual sum processing on the input vector of the residual block and the second convolution vector through the residual sum layer to obtain an output vector includes:

[0045] Through the residual sum layer, residual sum processing is performed on the input vector of the residual block and the second activated feature vector to obtain an output vector.

[0046] According to some embodiments of the first aspect of the present application, the training step of the parking space recommendation model includes:

[0047] Acquire historical parking data, wherein the historical parking data includes user rating data;

[0048] The historical parking data is filtered based on the user rating data to obtain target historical data; the target historical data includes multiple training data, each of which includes historical user information, historical time information, historical user target location information, historical user location information, historical intended parking duration, multiple historical candidate parking spaces, selected parking spaces, and historical parking space price information of the historical candidate parking spaces;

[0049] Constructing a training vector to be extracted based on the historical user information, the historical time information, the historical parking space price information, the historical intended parking duration, the historical user target location information, and the historical user location information;

[0050] Inputting the to-be-extracted training vector into an initial parking space recommendation model;

[0051] Performing residual feature extraction on the training vector to be extracted by using a plurality of the residual blocks to obtain a residual training feature vector;

[0052] Inputting the residual training feature vector into the long short-term memory module to obtain an updated training feature vector;

[0053] By means of the fully connected layer, the updated training feature vector is associated with each of the historical candidate parking spaces to obtain a plurality of associated training feature vectors;

[0054] Inputting a plurality of the associated training feature vectors into the classifier to obtain a training probability distribution of the associated training feature vectors;

[0055] Based on the training probability distribution, taking the historical candidate parking space corresponding to the associated training feature vector with the largest probability value as the target training parking space;

[0056] A loss value is calculated based on the target training parking space and the selected parking space, and the initial parking space recommendation model is iteratively updated based on the loss value to obtain a trained parking space recommendation model.

[0057] According to some embodiments of the first aspect of the present application, sending the relevant information of the target parking space to the user terminal includes:

[0058] Based on the target parking space, obtaining target parking space location information and price information of the target parking space;

[0059] The target parking space location information and price information are sent to the user terminal.

[0060] A second aspect of the present application provides a parking space recommendation device, including:

[0061] a receiving module, configured to, in response to receiving a parking space search request instruction sent by a user terminal, determine parking demand information based on the parking space search request instruction, wherein the parking demand information includes user information, current time information, user target location information, user current location information, intended parking duration, multiple current candidate parking spaces, and parking space price information of the current candidate parking spaces;

[0062] A vector construction module, used to construct a vector to be extracted based on the user information, the current time information, the planned parking duration, the parking space price information, the user target location information and the user current location information;

[0063] An input module, used for inputting the vector to be extracted into a trained parking space recommendation model; the parking space recommendation model comprises a long short-term memory module, a fully connected layer, a classifier and a plurality of residual blocks connected in sequence;

[0064] An extraction module, used for performing residual feature extraction on the vector to be extracted through a plurality of the residual blocks to obtain a residual feature vector;

[0065] An updating module, used for inputting the residual feature vector into the long short-term memory module to obtain an updated feature vector;

[0066] An association module, used to associate the updated feature vector with each of the current candidate parking spaces through the fully connected layer to obtain a plurality of associated feature vectors;

[0067] A classification module, used for inputting a plurality of the associated feature vectors into the classifier to obtain a probability distribution of the associated feature vectors;

[0068] A selection module, configured to select the current candidate parking space corresponding to the associated feature vector with the largest probability value as the target parking space based on the probability distribution;

[0069] The sending module is used to send the relevant information of the target parking space to the user terminal.

[0070] A third aspect of the present application provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the parking space recommendation method described in any one of the first aspect embodiments is implemented.

[0071] The fourth aspect of the present application provides a computer-readable storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by a processor, the parking space recommendation method described in any one of the first aspect embodiments is implemented.

[0072] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] The present application is further described below with reference to the accompanying drawings and embodiments, wherein:

[0074] Figure 1A schematic diagram of the structure of a parking control system according to an embodiment of the present application;

[0075] Figure 2 A schematic diagram of the steps of the parking space recommendation method according to an embodiment of the present application;

[0076] Figure 3 A schematic diagram of an interface in a parking program in an embodiment of the present application;

[0077] Figure 4 A schematic diagram of the structure of a parking space recommendation model according to an embodiment of the present application;

[0078] Figure 5 A schematic diagram of the specific structure of the residual block in an embodiment of the present application;

[0079] Figure 6 This is a schematic diagram of the specific structure of the extended causal convolutional layer of an embodiment of the present application;

[0080] Figure 7 A schematic diagram of the structure of a parking space recommendation device according to an embodiment of the present application;

[0081] Figure 8 A schematic diagram of the structure of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0082] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be understood as limiting the present application.

[0083] In the description of the present application, it should be understood that descriptions involving orientation, such as up, down, front, back, left, right, etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present application.

[0084] In the description of this application, "several" means more than one, "more" means more than two, "greater than", "less than", "exceed", etc. are understood to exclude the number itself, and "above", "below", "within", etc. are understood to include the number itself. If there is a description of "first" or "second", it is only used for the purpose of distinguishing technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features or implicitly indicating the order of the indicated technical features.

[0085] In the description of this application, unless otherwise clearly defined, terms such as setting, installing, connecting, etc. should be understood in a broad sense, and technicians in the relevant technical field can reasonably determine the specific meanings of the above terms in this application based on the specific content of the technical solution.

[0086] In the description of the present application, the description with reference to the terms "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0087] The parking space recommendation method provided in the embodiment of the present application relates to the field of artificial intelligence technology. The parking space recommendation method provided in the embodiment of the present application can be applied to a terminal, can be applied to a server side, or can be software running in a terminal or a server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application that implements the parking space recommendation method, etc., but is not limited to the above forms.

[0088] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0089] The present application embodiment provides a parking control system, referring to Figure 1 , Figure 1 This is a schematic diagram of the structure of the parking control system of the embodiment of the present application. The parking control system is used to execute the parking space recommendation method of the embodiment of the present application. The parking control system includes a main server and multiple slave servers. The server is communicatively connected to the multiple slave servers, and the main server is communicatively connected to the user terminal. Each slave server corresponds to a parking lot. Each parking lot is provided with a visual device. The visual device is communicatively connected to the corresponding slave server. The parking spaces are monitored through the visual device to collect relevant information. For example, the vacant parking spaces in the parking lot at a certain moment are collected through the visual device. For example, the parking time of a certain car is collected through the visual device. In each slave server, there is also a database, which stores user information, parking price information and historical data. The trained parking space recommendation model is configured in the main server.

[0090] It should be noted that the user terminal is a device used to send parking space search request instructions to the main server. The user terminal can be in various forms such as laptops, PDAs (personal digital assistants), mobile phones, vehicle-mounted terminals, and dedicated terminals. In addition, it can be a single device or a combination of multiple devices. For example, a mobile phone and a tablet computer are connected to each other through a local area network, share a display, etc. to work together, and together constitute a user terminal. The terminal can communicate with the Internet in a wired or wireless manner, and then exchange data with the main server.

[0091] It should be noted that a server refers to a computer system that can provide certain services to user terminals. Compared with ordinary terminals, servers have higher requirements in terms of stability, security, performance, etc. A server can be a high-performance computer in a network platform, a cluster of multiple high-performance computers, a part of a high-performance computer (such as a virtual machine), a combination of parts of multiple high-performance computers (such as virtual machines), etc.

[0092] It should be noted that in each specific implementation of the present application, when it comes to the need to perform relevant processing based on data related to user identity or characteristics such as user information, user behavior data, user historical data, and user location information, the user's permission or consent will be obtained first. For example, when obtaining user stored data and the user's cached data access request, the user's permission or consent will be obtained first; when obtaining resource data, the present application embodiment will first obtain the user's permission or consent. Moreover, the collection, use and processing of these data will comply with relevant laws, regulations and standards. In addition, when the present application embodiment needs to obtain the user's sensitive personal information, the user's separate permission or consent will be obtained through a pop-up window or by jumping to a confirmation page. After clearly obtaining the user's separate permission or consent, the necessary user-related data used to enable the normal operation of the present application embodiment will be obtained.

[0093] based on Figure 1 A parking space recommendation method according to the first aspect of the present application is proposed. Figure 2 , Figure 2 The following is a flow chart of the steps of the parking space recommendation method according to an embodiment of the present application. The parking space recommendation method includes but is not limited to the following steps:

[0094] Step S210, in response to receiving a parking space search request instruction sent by a user terminal, determining parking demand information based on the parking space search request instruction, the parking demand information including user information, current time information, user target location information, user current location information, intended parking duration, multiple current candidate parking spaces, and parking space price information of the current candidate parking spaces;

[0095] It is worth noting that a parking program is installed in the user terminal. The user logs in to the pre-registered account in the parking program. When parking is needed, the parking program is opened, and the parking program is controlled by the buttons or touch screen of the terminal to send a parking space search request instruction to the main server. The parking space search instruction includes some parking demand information. In the parking demand information, the user information is the user's unique code UID in the parking program, and the UID uniquely indicates the user. The current time information represents the moment when the user terminal sends the parking space search instruction; the user's target location information refers to the user's destination; and the user's current location information refers to the current location of the user's vehicle; the intended parking time refers to the time the user expects to need to park; the candidate parking space refers to the currently vacant parking space; the parking space price information of the current candidate parking space refers to the parking space price information, such as a vacant parking space in a parking lot, and the parking price is 20 yuan / hour. It should be noted that multiple current candidate parking spaces can be parking spaces in different parking lots.

[0096] For example, the parking program displays an interface to the user, and guides the user to prompt relevant information in a text box of the interface through interface elements, so as to obtain parking demand information. Figure 3 , Figure 3 This is a schematic diagram of the interface in the parking program of the embodiment of the present application. Figure 3 In the text box to the right of "Please fill in the destination" is used by the user to fill in the user's target location information; the text box to the right of "Please fill in the parking duration" is used by the user to fill in the intended parking duration. After the user completes the filling, click the "Send" button to send a parking space search request instruction to the main server. The user's current location information is obtained through the GPS positioning module of the user terminal; multiple current candidate parking spaces and the parking space price information of the current candidate parking spaces are obtained through the main server and multiple slave servers.

[0097] Step S220, constructing a vector to be extracted based on user information, current time information, planned parking duration, parking space price information, user target location information, and user current location information;

[0098] It is worth noting that since user information, current time information, planned parking time, parking space price information, user target location information and user current location information cannot be directly used as input of the neural network model, it is necessary to convert user information, current time information, planned parking time, parking space price information, user target location information and user current location information into sequence vectors, construct the vector to be extracted, and then input the vector to be extracted into the parking space recommendation model.

[0099] Step S230, inputting the vector to be extracted into a trained parking space recommendation model; the parking space recommendation model includes a long short-term memory module, a fully connected layer, a classifier, and a plurality of residual blocks connected in sequence;

[0100] Specifically, refer to Figure 4 , Figure 4 The structure diagram of the parking space recommendation model of the embodiment of the present application is shown in FIG. The vector to be extracted is first input to a plurality of residual blocks connected in sequence, and then processed by the plurality of residual blocks, and then sequentially input to the long short-term memory module, the fully connected layer and the classifier. Figure 4 In the embodiment, the number of residual blocks is 4, which is only an example and cannot be understood as a limitation on the number of residual blocks. Those skilled in the art can set the number of residual blocks according to actual conditions.

[0101] It is worth noting that the Long Short-Term Memory (LSTM) module is a special recurrent neural network (RNN) designed to solve the long-term dependency problem encountered by RNN when processing long sequence data. LSTM is a time recurrent neural network that, through its unique structural design, can remember values ​​of indefinite time lengths and effectively handle long-term dependencies in time series data. LSTM controls the storage and deletion of information by introducing three gating mechanisms (forget gate, input gate, output gate) and a memory cell (also called cell state). These gating mechanisms are composed of sigmoid functions and tanh functions, which calculate a real number vector between 0 and 1 to determine the degree of information passing. LSTM effectively solves the gradient vanishing or exploding problem encountered by RNN when processing long sequence data through the design of gating mechanisms and cell states, thereby being able to capture long-term dependencies in the sequence.

[0102] It should be noted that although the vector to be extracted in the embodiment of the present application is a non-time series, since the vector to be extracted in the embodiment of the present application includes more information (user information, current time information, planned parking time, parking space price information, user target location information and user current location information), these information have contextual information and dependencies. The vector to be extracted can be regarded as a long time series. Therefore, the vector to be extracted in the embodiment of the present application is suitable for regression prediction through LSTM, and the dependency relationship between user information, current time information, planned parking time, parking space price information, user target location information and user current location information can be captured through LSTM.

[0103] Step S240, performing residual feature extraction on the vector to be extracted through multiple residual blocks to obtain a residual feature vector;

[0104] Step S250, inputting the residual feature vector into the long short-term memory module to obtain an updated feature vector;

[0105] Step S260, through the fully connected layer, the updated feature vector is associated with each current candidate parking space to obtain multiple associated feature vectors;

[0106] In one embodiment, in the fully connected layer, the process of performing the association is expressed as follows:

[0107] ;

[0108] in, Represents the updated feature vector, The associated features that represent the association between the updated feature vector and the i-th current candidate parking space, and fully_connected represents the function of the fully connected mapping, which is used to map the updated feature vector with the current candidate parking space.

[0109] In one embodiment, the updated feature vector is [a0, a1, a2, a3...an]. The parking control system includes a total of multiple parking lots, and the parking spaces of the parking lot are represented as Xij, i represents the i-th parking lot, and j represents the j-th parking space. Each parking space representation of each parking lot is encoded, for example, using one-hot encoding, so as to convert each parking space representation into a vector. The current candidate parking spaces are X12, X23, and X45, respectively, and their encoded vectors are A, B, and C, respectively. The function of the fully connected mapping can be a splicing function, then in the fully connected layer, the updated feature vector is spliced ​​with the vector of each current candidate parking space respectively to obtain multiple associated features. The associated features obtained after associating the updated feature vector with X12 are [a0, a1, a2, a3...an, A]; the associated features obtained after associating the updated feature vector with X23 are [a0, a1, a2, a3...an, B]; the associated features obtained after associating the updated feature vector with X45 are [a0, a1, a2, a3...an, C].

[0110] Step S270, inputting the plurality of associated feature vectors into a classifier to obtain a probability distribution of the associated feature vectors;

[0111] In one embodiment, in one embodiment, a Softmax function is used as a classifier. The Softmax function can convert the original output into a probability distribution so that the output value of each category is between 0 and 1, and the sum of the probabilities of all categories is 1. Compared with other normalization methods, such as simple maximum value normalization, the Softmax function is more stable in terms of value. When the input value is large or small, the Softmax function can avoid the problem of numerical overflow or underflow, thereby ensuring the stability and accuracy of the model. The calculation of the Softmax function is relatively simple and can be efficiently calculated in a vectorized manner. This makes it possible to reduce computing time and resource consumption when training large-scale data sets.

[0112] Step S280, based on the probability distribution, taking the current candidate parking space corresponding to the associated feature vector with the largest probability value as the target parking space;

[0113] Step S290: Send relevant information of the target parking space to the user terminal.

[0114] According to the parking space recommendation method of the embodiment of the present application, through the above steps S210 to S290, when receiving the parking space search request instruction sent by the user terminal, the parking demand information is determined based on the parking space search request instruction, and then the vector to be extracted is constructed based on the parking demand information, and the vector to be extracted is input into the parking space recommendation model. In the parking space recommendation model, residual features are extracted from the vector to be extracted through multiple residual blocks. The residual blocks can effectively alleviate the problems of gradient disappearance and gradient explosion, and multiple residual blocks are connected in sequence, which can increase the depth of the parking space recommendation model, so as to help the parking space recommendation model better utilize context information, thereby improving the expression ability and performance of the parking space recommendation model. Then, the residual feature vector is updated through the long short-term memory module, and the dependency between each piece of information in the parking demand information can be captured through the long short-term memory module, so as to further improve the performance of the parking space recommendation model, improve the prediction ability of the parking space recommendation model, and help the parking space recommendation model predict the target parking space that meets the user's needs. After obtaining the updated feature vector through the long short-term memory module, the updated feature vector is associated with each current candidate parking space through the fully connected layer to obtain multiple associated feature vectors; the multiple associated feature vectors are input into the classifier to obtain the probability distribution of the associated feature vectors; based on the probability distribution, the current candidate parking space corresponding to the associated feature vector with the largest probability value is used as the target parking space, and the relevant information of the target parking space is sent to the user terminal. In this way, the present application realizes the automatic recommendation of parking spaces that meet the user's needs based on the parking demand information.

[0115] In one embodiment, referring to Figure 5 , Figure 5 Schematic diagram of the specific structure of the residual block of the embodiment of the present application. The residual block includes a first extended causal convolution layer, a channel attention module, a spatial attention module, a multiplier, a second extended causal convolution layer and a residual sum layer. Figure 5 In , the global attention module consists of a channel attention module and a spatial attention module. Figure 6 , Figure 6 This is a schematic diagram of the specific structure of the extended causal convolutional layer of an embodiment of the present application. Figure 6 The illustrated structure may be a first extended causal convolution layer or a second extended causal convolution layer, and the first extended causal convolution layer has the same structure as the second extended causal convolution layer.

[0116] In the extended causal convolution layer, padding the sequence by the padding coefficient can ensure that the output sequence has the same length as the input sequence, and can expand the receptive field so that the convolution kernel can capture a wider range of contextual information. For example, a padding coefficient = 2 means adding 2 units of padding values ​​on both sides of the input sequence (usually the left and right sides). These padding values ​​are usually set to 0, so they are also called zero padding. If the length of the input sequence is L, then after applying the padding coefficient = 2, the length of the sequence will become L+2*2=L+4. In causal convolution, since the output of each time step can only depend on current and past information, if padding is not used, some input information may be ignored during the convolution process. By adding padding, it can be ensured that all information in the input sequence is taken into account by the convolution kernel, thereby avoiding information loss. Therefore, in the extended causal convolution layer of the present application, padding the sequence by the padding coefficient can avoid information loss of the input vector. d is the expansion coefficient. The main function of the expansion coefficient is to increase the receptive field of the convolution layer, that is, the time range of the input sequence that the network can capture. By applying different expansion coefficients in the convolution operations of different layers, the model's field of view can be effectively expanded. The expansion coefficient determines the interval between adjacent elements in the convolution kernel. When the expansion coefficient is 1, the convolution kernel is regular; when the expansion coefficient is greater than 1, the convolution kernel will skip some elements in the input sequence for convolution operations. This design enables high-level network models to capture information over a longer time span, which helps to process long-term dependencies in sequence data. The exponential growth of the expansion coefficient of the embodiment of the present application can ensure that the network's receptive field (i.e., the length of the input sequence that the network can cover) can expand rapidly as the number of layers increases. This is very important for processing long sequence data, because it ensures that the parking space recommendation model can capture dependencies over a longer time range. Therefore, the parking space recommendation model of the present application can capture dependencies between parking demand information.

[0117] It is worth noting that the input vector of the residual block at the head of the residual blocks connected in sequence is the vector to be extracted;

[0118] The output vector of the last residual block among the sequentially connected residual blocks is used as the residual feature vector;

[0119] The processing process of each residual block is:

[0120] Step S241, inputting the input vector of the residual block into the first extended causal convolution layer to obtain a first convolution vector;

[0121] Step S242, inputting the first convolution vector into a channel attention module to obtain a channel attention feature vector;

[0122] Step S243, inputting the channel attention feature vector into the spatial attention module to obtain the spatial attention feature vector;

[0123] Step S244, inputting the first convolution vector and the spatial attention feature vector into a multiplier to obtain a multiplication vector;

[0124] It is worth noting that the first convolution vector and the spatial attention feature vector are multiplied by a multiplier to obtain a multiplication vector.

[0125] Step S245, inputting the multiplication vector into the second extended causal convolution layer to obtain a second convolution vector;

[0126] Step S246, performing residual sum processing on the input vector of the residual block and the second convolution vector through the residual sum layer to obtain an output vector.

[0127] In one embodiment, referring to Figure 5 , the residual block also includes a first weight normalization layer and a first rectified linear unit.

[0128] Before step S242, the following steps may be included:

[0129] Step S2421, performing a weight normalization operation on the first convolution vector through a first weight normalization layer to obtain a first weight normalized feature vector;

[0130] Step S2422, performing nonlinear activation on the first weighted normalized eigenvector by using a first rectified linear unit to obtain a first activated eigenvector;

[0131] Step S242 includes:

[0132] Step S2422: input the first activated feature vector into the channel attention module to obtain a channel attention feature vector.

[0133] In one embodiment, referring to Figure 5 , the residual block also includes a second weight normalization layer and a second rectified linear unit;

[0134] Before step S246, the following steps may be included:

[0135] Step S2461, performing a weight normalization operation on the second convolution vector through a second weight normalization layer to obtain a second weight normalized feature vector;

[0136] Step S2462, performing nonlinear activation on the second weighted normalized eigenvector by a second rectified linear unit to obtain a second activated eigenvector;

[0137] Step S246 includes:

[0138] Step S2463, through the residual sum layer, the input vector of the residual block and the second activated feature vector are subjected to residual sum processing to obtain an output vector.

[0139] It is worth noting that in the parking space recommendation model, as the number of network model layers increases, the gradient may gradually disappear or explode during the back propagation process, making the network difficult to train. The residual connection creates a direct path from input to output, making it easier for the gradient to pass through the entire network during the propagation process, thereby effectively alleviating the problem of gradient vanishing and gradient exploding. And because the residual connection allows the gradient to be more directly passed back to the previous layer, the model can converge to the optimal solution faster during the training process, improving training efficiency and reducing training time. Therefore, the embodiment of the present application performs a 1*1 convolution on the input vector of the residual block through the residual sum layer, and then performs residual sum processing on the input vector after the 1*1 convolution and the second activated feature vector to obtain the output vector of the residual block, which can effectively alleviate the problem of gradient vanishing and gradient exploding in the parking space recommendation model, and can improve training efficiency and reduce training time.

[0140] It is worth noting that the second weight normalization layer can be the same as the first weight normalization layer; the second rectified linear unit can be the same as the first rectified linear unit.

[0141] It is worth noting that the weight normalization operation can make each node contribute the same to the model, thereby improving the training efficiency and accuracy of the neural network. This is because normalization can reduce the numerical differences between different features, thereby speeding up the convergence of the gradient descent algorithm. Since the vector to be extracted can be regarded as a long time series, the first convolution vector is also equivalent to a long time series. Since it involves the processing of time series data, the eigenvalues ​​of different time steps may have large differences. Through weight normalization, these eigenvalues ​​can have a more equal status in the model, thereby improving the performance of the parking space recommendation model.

[0142] It is worth noting that the nonlinear activation function can introduce nonlinear factors, enabling the neural network to learn and fit complex functions. Without a nonlinear activation function, no matter how many layers the neural network has, it can only calculate linear functions, which greatly limits the expressive power of the neural network. The present application performs nonlinear activation on the first weighted normalized eigenvector through a first corrected linear unit, which can improve the expressive power of the parking space recommendation model.

[0143] In one embodiment, in step S2422, the first activation feature vector is subjected to channel attention processing by the channel attention module to obtain a channel attention coefficient, and then the first activation feature vector is multiplied by the channel attention coefficient to obtain a channel attention feature vector. The processing process is expressed as:

[0144] ;

[0145] Among them, Mc(F1) represents the channel attention coefficient obtained by performing channel attention processing on the first activated feature vector; F1 represents the first activated feature vector; F2 represents the channel attention feature vector.

[0146] In one embodiment, in step S243, the spatial attention module performs spatial attention processing on the channel attention feature vector to obtain a spatial attention coefficient, and then multiplies the spatial attention coefficient by the channel attention feature vector to obtain a spatial attention feature vector. The processing process can be expressed as:

[0147] ;

[0148] Among them, Ms(F2) represents the spatial attention coefficient obtained by performing spatial attention processing on the channel attention feature vector; F2 represents the channel attention feature vector; F3 represents the spatial attention feature vector.

[0149] In one embodiment, step S220 includes the following steps:

[0150] Step S221, converting the unique code in the user information into a first array;

[0151] In one embodiment, the embodiment of the present application uses UID as a unique code, and UID is a string consisting of multiple digits, and the string is converted into binary data to obtain a first array. For example, if UID is "212457456", it is converted into binary data: 1100101010011101011111110000, and the first array is [1100101010011101011111110000].

[0152] Step S222, converting the coordinate values ​​in the user target position information into a second array;

[0153] In one embodiment, the coordinate value of the user target position is determined from the user target position information, the coordinate value is divided into longitude and latitude, the longitude and latitude are converted into binary data respectively, and the binary data of the longitude and the binary data of the latitude are spliced ​​to obtain a second array. Exemplarily, the longitude and latitude of a user target position are 30.526 and 90.546 respectively, then 30.526 is first removed from the decimal point to obtain 30526, and 30526 is converted into binary data: 111011100111110; 90.546 is removed from the decimal point to obtain 90546, and 90546 is converted into binary data: 10110000110110010, then the second array is [111011100111110, 10110000110110010].

[0154] Step S223, converting the coordinate values ​​in the user's current location information into a third array;

[0155] It should be noted that the conversion method of step S223 is the same as that of step S222, and will not be repeated here.

[0156] Step S224, converting the parking space price information into a fourth array;

[0157] In one embodiment, the parking space price information includes parking lot charging rules. The hourly price of the parking space is first calculated based on the parking lot charging rules and the planned parking duration, and then the hourly price is converted into binary data as the fourth array.

[0158] In one embodiment, if the parking lot charging rule is that the parking space price is X / hour, then X is the hourly price of the parking space. For example, if X is 20, then 20 is converted into binary data 10100, and the fourth array is

[10100] .

[0159] In another embodiment, the parking lot charging rule is stage charging, specifically: if the parking time is less than or equal to A, then Y yuan per hour; if the parking time is greater than A, then Z yuan per hour. In this case, if the intended parking time is less than or equal to A, Y is used as the hourly price; if the intended parking time is greater than A, Z is used as the hourly price.

[0160] Step S225, converting the current time information into a fifth array;

[0161] In one embodiment, the current time information is 12:00, and 1200 is converted into binary data as 10010110000. Then the fifth array is [10010110000].

[0162] Step S226, converting the planned parking duration into a sixth array;

[0163] In one embodiment, the planned parking time is 3 hours, that is, 180 minutes, and 180 is converted into binary data as 10110100. Therefore, the sixth array is [10110100].

[0164] Step S227, obtaining a vector to be extracted based on the first array, the second array, the third array, the fourth array, the fifth array and the sixth array.

[0165] In step S227, the first array, the second array, the third array, the fourth array, the fifth array and the sixth array are concatenated in sequence, and the obtained arrays are used as the vectors to be extracted.

[0166] In one embodiment, the steps of training the parking space recommendation model include:

[0167] Step S310, obtaining historical parking data, the historical parking data including user rating data;

[0168] Step S320, filtering the historical parking data based on the user rating data to obtain target historical data; the target historical data includes multiple training data, each training data represents a parking record, and each training data includes user rating data, historical user information, historical time information, historical user target location information, historical user location information, historical intended parking duration, multiple historical candidate parking spaces, selected parking spaces, and historical parking space price information of historical candidate parking spaces;

[0169] Specifically, when it is detected that the user's vehicle has left the parking lot, a rating interface is sent to the user terminal, and the user controls the rating interface through the terminal to rate the parking experience and obtain user rating data. Parking records with user rating data below the rating threshold are removed, and parking records with user rating data above or equal to the rating threshold are retained to obtain target historical data. Therefore, the target historical data is the data of user satisfaction parking experience.

[0170] Step S330, constructing a training vector to be extracted based on historical user information, historical time information, historical parking space price information, historical intended parking duration, historical user target location information, and historical user location information;

[0171] It should be noted that historical user information refers to the information of a certain user in the historical parking data, historical time information refers to the parking time of the user; historical parking space price information refers to the parking space price information of the user in the historical time information; historical reverse parking duration is the intended parking duration of the user in the historical time information, historical user target location information is the user target location information of the user in the historical time information; historical user location information is the user location information of the user in the historical time information. Historical candidate parking spaces refer to parking spaces that are vacant in the historical time information. The selected parking space is the parking space selected by the user from multiple historical candidate parking spaces.

[0172] It should be noted that the process of constructing the training vector to be extracted is the same as the process of constructing the vector to be extracted.

[0173] Step S340, inputting the training vector to be extracted into the initial parking space recommendation model;

[0174] Step S350, performing residual feature extraction on the training vector to be extracted through multiple residual blocks to obtain a residual training feature vector;

[0175] Step S360, inputting the residual training feature vector into the long short-term memory module to obtain an updated training feature vector;

[0176] Step S370, through the fully connected layer, the updated training feature vector is associated with each historical candidate parking space to obtain multiple associated training feature vectors;

[0177] Step S380, inputting the plurality of associated training feature vectors into a classifier to obtain a training probability distribution of the associated training feature vectors;

[0178] Step S390, based on the training probability distribution, taking the historical candidate parking space corresponding to the associated training feature vector with the largest probability value as the target training parking space;

[0179] It should be noted that the processing process of the training vector to be extracted in the parking space recommendation model is the same as the processing process of the vector to be extracted in the parking space recommendation model, which will not be repeated here.

[0180] Step S400, calculating the loss value based on the target training parking space and the selected parking space, and iteratively updating the initial parking space recommendation model based on the loss value to obtain a trained parking space recommendation model.

[0181] It should be noted that after obtaining the target training parking spaces and the selected parking spaces, the preset loss function is used to calculate the loss value. After obtaining the loss value, the initial parking space recommendation model is updated based on the loss value. This cycle is repeated until the number of cycles reaches the preset number, thereby obtaining a trained parking space recommendation model.

[0182] Specifically, the target training parking space and the selected parking space are converted into vectors respectively, for example, one-hot encoding is used to represent the vectors of the target training parking space and the selected parking space respectively, and then the loss value between the vectors of the target training parking space and the selected parking space is calculated by the loss function. Loss value is a very important concept in machine learning and deep learning. It is used to quantify the difference or error between the model prediction results and the actual observation results. This application does not limit the loss function, for example, it can be a cross entropy function.

[0183] The embodiment of the present application performs iterative training and optimization on the initial parking space recommendation model based on the target historical data through steps S310 to S400, thereby obtaining a trained parking space recommendation model. Since the historical parking data is screened based on the user rating data to obtain the target historical data, the parking spaces in the target historical data are all parking spaces that the user is relatively satisfied with, so that the trained parking space recommendation model can automatically recommend parking spaces that meet the user's needs to the user based on the parking demand information.

[0184] In one embodiment, step S290 may include the following steps:

[0185] Step S291, based on the target parking space, obtaining target parking space location information and price information of the target parking space;

[0186] Step S292: Send the target parking space location information and price information to the user terminal.

[0187] The second aspect of the present application provides a parking space recommendation device, referring to Figure 7 , Figure 7 This is a schematic diagram of the structure of the parking space recommendation device of the embodiment of the present application. The device includes:

[0188] The receiving module 710 is used to determine parking demand information based on the parking space search request instruction in response to receiving a parking space search request instruction sent by the user terminal, where the parking demand information includes user information, current time information, user target location information, user current location information, intended parking duration, multiple current candidate parking spaces, and parking space price information of the current candidate parking spaces;

[0189] A vector construction module 720 is used to construct a vector to be extracted based on user information, current time information, planned parking duration, parking space price information, user target location information and user current location information;

[0190] An input module 730 is used to input the vector to be extracted into a trained parking space recommendation model; the parking space recommendation model includes a long short-term memory module, a fully connected layer, a classifier, and a plurality of residual blocks connected in sequence;

[0191] An extraction module 740 is used to perform residual feature extraction on the vector to be extracted through multiple residual blocks to obtain a residual feature vector;

[0192] An updating module 750, configured to input the residual feature vector into the long short-term memory module to obtain an updated feature vector;

[0193] An association module 760 is used to associate the updated feature vector with each current candidate parking space through a fully connected layer to obtain multiple associated feature vectors;

[0194] A classification module 770, for inputting a plurality of associated feature vectors into a classifier to obtain a probability distribution of the associated feature vectors;

[0195] A selection module 780 is used to select the current candidate parking space corresponding to the associated feature vector with the largest probability value as the target parking space based on the probability distribution;

[0196] The sending module 790 is used to send relevant information of the target parking space to the user terminal.

[0197] The parking space recommendation device of the embodiment of the present application is used to execute the parking space recommendation method of the embodiment of the first aspect of the present application. When executing the method, upon receiving the parking space search request instruction sent by the user terminal, the parking demand information is determined based on the parking space search request instruction, and then a vector to be extracted is constructed based on the parking demand information, and the vector to be extracted is input into the parking space recommendation model. In the parking space recommendation model, residual features are extracted from the vector to be extracted through multiple residual blocks. The residual blocks can effectively alleviate the problems of gradient disappearance and gradient explosion, and multiple residual blocks are connected in sequence, which can increase the depth of the parking space recommendation model, so as to help the parking space recommendation model better utilize context information, thereby improving the expression ability and performance of the parking space recommendation model. Then, the residual feature vector is updated through the long short-term memory module, and the dependency between each piece of information in the parking demand information can be captured through the long short-term memory module, so as to further improve the performance of the parking space recommendation model, improve the prediction ability of the parking space recommendation model, and help the parking space recommendation model predict the target parking space that meets the user's needs. After obtaining the updated feature vector through the long short-term memory module, the updated feature vector is associated with each current candidate parking space through the fully connected layer to obtain multiple associated feature vectors; the multiple associated feature vectors are input into the classifier to obtain the probability distribution of the associated feature vectors; based on the probability distribution, the current candidate parking space corresponding to the associated feature vector with the largest probability value is used as the target parking space, and the relevant information of the target parking space is sent to the user terminal. In this way, the present application realizes the automatic recommendation of parking spaces that meet the user's needs based on the parking demand information.

[0198] It should be noted that the specific implementation of the parking space recommendation device is basically the same as the specific implementation of the above-mentioned parking space recommendation method, and will not be repeated here. On the premise of meeting the requirements of the embodiment of the present application, the parking space recommendation device can also be provided with other functional units to implement the parking space recommendation method in the above-mentioned embodiment.

[0199] The third aspect of the present application provides an electronic device, the electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements any parking space recommendation method of the first aspect when executing the computer program. The electronic device can be any smart terminal including a tablet computer, a car computer, etc.

[0200] Reference Figure 8 , Figure 8 FIG. 1 is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. The electronic device includes:

[0201] The processor 801 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0202] The memory 802 may be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 802 may store an operating system and other application programs. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 802, and the processor 801 calls and executes the parking space recommendation method of the embodiment of this application;

[0203] Input / output interface 803, used to implement information input and output;

[0204] The communication interface 804 is used to realize the communication interaction between the device and other devices. The communication can be realized through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.);

[0205] A bus 805 that transmits information between various components of the device (e.g., the processor 801, the memory 802, the input / output interface 803, and the communication interface 804);

[0206] The processor 801 , the memory 802 , the input / output interface 803 and the communication interface 804 are connected to each other in communication within the device via a bus 805 .

[0207] The fourth aspect of the present application is a computer-readable storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by a processor, the parking space recommendation method of any one of the first aspect embodiments is implemented.

[0208] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor 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.

[0209] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0210] Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0211] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0212] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0213] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0214] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the mapping relationship of the mapping objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the previous and next mapping objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0215] In the several embodiments provided in the present application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0216] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0217] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0218] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including multiple instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs.

[0219] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present invention should be within the scope of the rights of the present invention.

Claims

1. A parking space recommendation method, characterized in that: include: In response to receiving a parking space search request instruction sent by a user terminal, determining parking demand information based on the parking space search request instruction, the parking demand information including user information, current time information, user target location information, user current location information, intended parking duration, multiple current candidate parking spaces, and parking space price information of the current candidate parking spaces; Constructing a vector to be extracted based on the user information, the current time information, the planned parking duration, the parking space price information, the user target location information and the user current location information; Inputting the vector to be extracted into a trained parking space recommendation model; the parking space recommendation model includes a long short-term memory module, a fully connected layer, a classifier, and a plurality of residual blocks connected in sequence; Performing residual feature extraction on the vector to be extracted by using a plurality of the residual blocks to obtain a residual feature vector; Inputting the residual feature vector into the long short-term memory module to obtain an updated feature vector; By means of the fully connected layer, the updated feature vector is associated with each of the current candidate parking spaces to obtain a plurality of associated feature vectors; Inputting a plurality of the associated feature vectors into the classifier to obtain a probability distribution of the associated feature vectors; Based on the probability distribution, taking the current candidate parking space corresponding to the associated feature vector with the largest probability value as the target parking space; Sending relevant information of the target parking space to the user terminal; The residual block includes a first extended causal convolution layer, a channel attention module, a spatial attention module, a multiplier, a second extended causal convolution layer, and a residual sum layer; The input vector of the residual block located at the head among the residual blocks connected in sequence is the vector to be extracted; The output vector of the residual block located at the end of the residual blocks connected in sequence is used as the residual feature vector; The processing process of each residual block is: Inputting the input vector of the residual block to the first extended causal convolution layer to obtain a first convolution vector; Inputting the first convolution vector into the channel attention module to obtain a channel attention feature vector; Inputting the channel attention feature vector into the spatial attention module to obtain a spatial attention feature vector; Inputting the first convolution vector and the spatial attention feature vector into the multiplier to obtain a multiplication vector; Inputting the multiplication vector into the second extended causal convolutional layer to obtain a second convolution vector; Through the residual sum layer, residual sum processing is performed on the input vector of the residual block and the second convolution vector to obtain an output vector.

2. The parking space recommendation method according to claim 1, characterized in that: Based on the user information, the current time information, the planned parking duration, the parking space price information, the user target location information and the user current location information, constructing a vector to be extracted, including: Converting the unique code in the user information into a first array; Converting the coordinate values ​​in the user target location information into a second array; Converting the coordinate values ​​in the user's current location information into a third array; Converting the parking space price information into a fourth array; Converting the current time information into a fifth array; Converting the planned parking duration into a sixth array; The vector to be extracted is obtained based on the first array, the second array, the third array, the fourth array, the fifth array and the sixth array.

3. The parking space recommendation method according to claim 1, characterized in that: The residual block also includes a first weight normalization layer and a first rectified linear unit; Before inputting the first convolution vector into the channel attention module, comprising: Performing a weight normalization operation on the first convolution vector through the first weight normalization layer to obtain a first weight normalized feature vector; performing nonlinear activation on the first weight-normalized feature vector by using the first rectified linear unit to obtain a first activated feature vector; The step of inputting the first convolution vector into the channel attention module to obtain a channel attention feature vector includes: The first activation feature vector is input into the channel attention module to obtain the channel attention feature vector.

4. The parking space recommendation method according to claim 1, characterized in that: The residual block also includes a second weight normalization layer and a second rectified linear unit; Before performing residual sum processing on the input vector of the residual block and the second convolution vector through the residual sum layer, the method further comprises: Performing a weight normalization operation on the second convolution vector through the second weight normalization layer to obtain a second weight normalized feature vector; performing nonlinear activation on the second weight-normalized feature vector by using the second rectified linear unit to obtain a second activated feature vector; The step of performing residual sum processing on the input vector of the residual block and the second convolution vector through the residual sum layer to obtain an output vector includes: Through the residual sum layer, residual sum processing is performed on the input vector of the residual block and the second activated feature vector to obtain an output vector.

5. The parking space recommendation method according to claim 1, characterized in that: The training steps of the parking space recommendation model include: Acquire historical parking data, wherein the historical parking data includes user rating data; The historical parking data is filtered based on the user rating data to obtain target historical data; the target historical data includes multiple training data, each of which includes historical user information, historical time information, historical user target location information, historical user location information, historical intended parking duration, multiple historical candidate parking spaces, selected parking spaces, and historical parking space price information of the historical candidate parking spaces; Constructing a training vector to be extracted based on the historical user information, the historical time information, the historical parking space price information, the historical intended parking duration, the historical user target location information, and the historical user location information; Inputting the to-be-extracted training vector into an initial parking space recommendation model; Performing residual feature extraction on the training vector to be extracted by using a plurality of the residual blocks to obtain a residual training feature vector; Inputting the residual training feature vector into the long short-term memory module to obtain an updated training feature vector; By means of the fully connected layer, the updated training feature vector is associated with each of the historical candidate parking spaces to obtain a plurality of associated training feature vectors; Inputting a plurality of the associated training feature vectors into the classifier to obtain a training probability distribution of the associated training feature vectors; Based on the training probability distribution, taking the historical candidate parking space corresponding to the associated training feature vector with the largest probability value as the target training parking space; A loss value is calculated based on the target training parking space and the selected parking space, and the initial parking space recommendation model is iteratively updated based on the loss value to obtain a trained parking space recommendation model.

6. The parking space recommendation method according to claim 1, characterized in that: The sending the relevant information of the target parking space to the user terminal includes: Based on the target parking space, obtaining target parking space location information and price information of the target parking space; The target parking space location information and price information are sent to the user terminal.

7. A parking space recommendation device, characterized in that: include: a receiving module, configured to, in response to receiving a parking space search request instruction sent by a user terminal, determine parking demand information based on the parking space search request instruction, wherein the parking demand information includes user information, current time information, user target location information, user current location information, intended parking duration, multiple current candidate parking spaces, and parking space price information of the current candidate parking spaces; A vector construction module, used to construct a vector to be extracted based on the user information, the current time information, the planned parking duration, the parking space price information, the user target location information and the user current location information; An input module, used for inputting the vector to be extracted into a trained parking space recommendation model; the parking space recommendation model comprises a long short-term memory module, a fully connected layer, a classifier and a plurality of residual blocks connected in sequence; An extraction module, used for performing residual feature extraction on the vector to be extracted through a plurality of the residual blocks to obtain a residual feature vector; An updating module, used for inputting the residual feature vector into the long short-term memory module to obtain an updated feature vector; An association module, used to associate the updated feature vector with each of the current candidate parking spaces through the fully connected layer to obtain a plurality of associated feature vectors; A classification module, used for inputting a plurality of the associated feature vectors into the classifier to obtain a probability distribution of the associated feature vectors; A selection module, configured to select the current candidate parking space corresponding to the associated feature vector with the largest probability value as the target parking space based on the probability distribution; A sending module, used for sending the relevant information of the target parking space to the user terminal; The residual block includes a first extended causal convolution layer, a channel attention module, a spatial attention module, a multiplier, a second extended causal convolution layer, and a residual sum layer; The input vector of the residual block located at the head among the residual blocks connected in sequence is the vector to be extracted; The output vector of the residual block located at the end of the residual blocks connected in sequence is used as the residual feature vector; The processing process of each residual block is: Inputting the input vector of the residual block to the first extended causal convolution layer to obtain a first convolution vector; Inputting the first convolution vector into the channel attention module to obtain a channel attention feature vector; Inputting the channel attention feature vector into the spatial attention module to obtain a spatial attention feature vector; Inputting the first convolution vector and the spatial attention feature vector into the multiplier to obtain a multiplication vector; Inputting the multiplication vector into the second extended causal convolutional layer to obtain a second convolution vector; Through the residual sum layer, residual sum processing is performed on the input vector of the residual block and the second convolution vector to obtain an output vector.

8. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the parking space recommendation method according to any one of claims 1 to 6 when executing the computer program.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the parking space recommendation method according to any one of claims 1 to 6 is implemented.

Images