CN113822458A

CN113822458A - Prediction method, training method, device, electronic equipment and readable storage medium

Info

Publication number: CN113822458A
Application number: CN202011291767.6A
Authority: CN
Inventors: 郜贺鹏; 郑宇�; 张钧波; 袁野
Original assignee: Jingdong City Beijing Digital Technology Co Ltd
Current assignee: Jingdong City Beijing Digital Technology Co Ltd
Priority date: 2020-11-18
Filing date: 2020-11-18
Publication date: 2021-12-21

Abstract

The disclosure provides a parking lot flow prediction method, a flow prediction model training method, a device, electronic equipment and a computer readable storage medium, and relates to the field of machine learning. The method for predicting the flow of the parking lot comprises the following steps: inputting isomorphic characteristics of a target parking lot into a first-layer model to output a description vector describing a flow trend and a first prediction vector, wherein the first-layer model is a language sequence translation model; inputting the description vector, the first prediction vector and heterogeneous features of the target parking lot into a second-layer model, and obtaining a second prediction vector according to the description vector and the heterogeneous features, wherein the second-layer model is a memory network model; and obtaining the predicted flow of the target parking lot based on the first prediction vector and the second prediction vector. Through the technical scheme disclosed by the invention, the traffic flow of the parking lot can be predicted on the basis of considering the type of the target parking lot, and the accuracy of the traffic flow prediction of the parking lot can be improved.

Description

Prediction method, training method, device, electronic equipment and readable storage medium

Technical Field

The present disclosure relates to the field of machine learning technologies, and in particular, to a method and an apparatus for predicting parking lot traffic, a method and an apparatus for training a traffic prediction model, an electronic device, and a computer-readable storage medium.

Background

In the construction of wisdom city, the parking area is indispensable building element, consequently makes the parking area maintain higher service level, is the key of guarantee traffic and peripheral attitude operation efficiency, makes things convenient for the resident trip. The traffic jam in the peripheral area of the parking lot can be relieved or solved, the scheduling efficiency of resources of the parking lot can be improved, and the operation difficulty of the parking lot is reduced.

The existing parking lot flow prediction scheme carries out flow prediction of a parking lot based on historical flow records of the parking lot, and has the following problems:

the existing parking lot flow prediction scheme is mainly realized based on the characteristics of inflow, outflow, parking saturation and the like, so that the prediction accuracy is poor.

It is to be noted that the information disclosed in the above background section is only for enhancement of understanding of the background of the present disclosure, and thus may include information that does not constitute prior art known to those of ordinary skill in the art.

Disclosure of Invention

The present disclosure is directed to a method for predicting parking lot traffic, a method and an apparatus for training a traffic prediction model, an electronic device, and a computer-readable storage medium, which overcome, at least to some extent, the problem of poor accuracy of parking lot traffic prediction in the related art.

Additional features and advantages of the disclosure will be set forth in the detailed description which follows, or in part will be obvious from the description, or may be learned by practice of the disclosure.

According to an aspect of the present disclosure, there is provided a method for predicting parking lot traffic, including: inputting time characteristics and corresponding flow characteristics of a target parking lot into a first-layer model to output a description vector and a first prediction vector for describing flow trend, wherein the first-layer model is a language sequence translation model; inputting the description vector, the first prediction vector and the heterogeneous features of the target parking lot into a second-layer model to obtain a second prediction vector according to the description vector and the heterogeneous features; and obtaining the predicted flow of the target parking lot according to the second prediction vector and the first prediction vector, wherein the second layer model is a memory network model.

In one embodiment, the memory network model includes an attention mechanism structure and a memory matrix, wherein the memory matrix is constructed based on different heterogeneous characteristics; the inputting the description vector, the first prediction vector and the heterogeneous features of the target parking lot into a second-layer model, and obtaining a second prediction vector according to the description vector and the heterogeneous features comprises: inputting the description vector and the heterogeneous features into the attention mechanism structure, and performing matching operation with the memory matrix; integrating results of the matching operations to generate an integrated vector, wherein the integrated vector describes the flow trend under the different heterogeneous characteristics; and obtaining the second prediction vector based on the integration vector and the heterogeneous characteristics of the target parking lot.

In one embodiment, further comprising: and configuring heterogeneous characteristics of the target parking lot based on the area type and/or the state type of the area where the target parking lot is located, wherein the area type comprises at least one of a transportation junction area, a business area, an office area and a residential area.

In one embodiment, the obtaining of the predicted traffic of the target parking lot based on the first prediction vector and the second prediction vector includes: performing fusion processing on the first prediction vector and the second prediction vector to obtain a fusion vector; inputting the fusion vector and the personality characteristics of the target parking lot into a third-layer model to perform full-connection processing on the fusion vector and the personality characteristics, wherein the third-layer model is a full-connection network model; and generating the predicted flow based on the processing result of the full-connection processing, and enabling the third-layer model to output the predicted flow.

In one embodiment, the obtaining of the predicted traffic of the target parking lot based on the first prediction vector and the second prediction vector includes: performing full-concatenation processing on the first prediction vector and the second prediction vector; and generating the predicted flow based on the processing result of the full-connection processing, and enabling the second layer model to output the predicted flow.

In one embodiment, the language order translation model comprises a Seq2Seq network, and the inputting isomorphic features of the target parking lot into the first layer model to output the description vector describing the traffic trend and the first prediction vector comprises: the isomorphic features comprise time features and corresponding flow features, and the time features and the flow features are preprocessed to generate corresponding preprocessing vectors; inputting the preprocessed vector into an encoder of the Seq2Seq network for encoding to obtain the description vector; and inputting the description vector and the prediction characteristics related to the prediction time into a decoder in the Seq2Seq network for decoding to obtain the first prediction vector.

In one embodiment, the inputting the time characteristic and the traffic characteristic of the target parking lot into the first layer model to output a description vector describing a traffic trend and a first prediction vector comprises: preprocessing the time characteristic and the flow characteristic to generate a corresponding preprocessing vector; inputting the preprocessed vector into an encoder of the Seq2Seq network for encoding to obtain the description vector; and inputting the description vector and the prediction characteristics related to the prediction time into a decoder in the Seq2Seq network for decoding to obtain the first prediction vector.

In one embodiment, the inputting the description vector and the prediction characteristics related to the prediction time into a decoder in the Seq2Seq network for decoding to obtain the first prediction vector further includes: determining the time characteristic of the prediction time and the prediction result of the previous moment as the prediction characteristic; and carrying out cooperative decoding on the description vector and the prediction characteristics to obtain the first prediction vector.

According to another aspect of the present disclosure, there is provided a method for training a flow prediction model, including: performing model training of a language sequence translation model based on historical data of a plurality of parking lots to generate a first-layer model, wherein the parking lots comprise target parking lots, and the historical data comprises isomorphic features of the parking lots; performing model training of a memory network on the output data of the first-layer model and heterogeneous characteristics of the target parking lot in a transfer learning mode to obtain a second-layer model; and generating a flow prediction model of the target parking lot based on the first layer model and the second layer model.

In one embodiment, the language sequence translation model comprises a Seq2Seq network, the model training of the language sequence translation model based on historical data of a plurality of parking lots, and the generating of the first layer model with the traffic prediction model comprises: the isomorphic characteristics of the parking lot comprise first isomorphic characteristics of the target parking lot and second isomorphic characteristics of other parking lots, and model training of the Seq2Seq network is carried out through the second isomorphic characteristics to obtain a basic model; and transferring the training result of the basic model to the first layer model to be trained, and performing model training on the first layer model to be trained through the first isomorphic feature to obtain the trained first layer model.

In one embodiment, the language sequence translation model comprises a Seq2Seq network, and the model training of the language sequence translation model based on historical data of a plurality of parking lots to generate the first layer model of the traffic prediction model comprises: the isomorphic characteristics of the parking lot comprise first isomorphic characteristics of the target parking lot and second isomorphic characteristics of other parking lots, and a plurality of training task sets are generated according to the second isomorphic characteristics; generating a meta-learning model of the Seq2Seq network according to the plurality of training task sets; and carrying out model learning on the meta-learning model based on the first isomorphic characteristic to obtain the first layer model.

In one embodiment, the performing, by using a transfer learning method, model training of a memory network on the output data of the first-layer model and the heterogeneous characteristics of the target parking lot to obtain a second-layer model includes: performing model training on the memory network through the training parameters of the first layer model to obtain structural parameters and a memory matrix of an attention structure in the memory network; and migrating the structural parameters and the memory matrix to the second-layer model to be trained, and performing model training on the second-layer model to be trained through the heterogeneous characteristics of the target parking lot to obtain the trained second-layer model.

In one embodiment, the generating a traffic prediction model for the target parking lot based on the first layer model and the second layer model comprises: performing model training of a fully-connected network on the output data of the first-layer model, the output data of the second-layer model and the individual characteristics of the target parking lot to obtain a third-layer model; generating the traffic prediction model based on the first layer model, the second layer model, and the third layer model.

According to still another aspect of the present disclosure, there is provided a prediction apparatus of parking lot traffic, including: the system comprises a first prediction module, a second prediction module and a third prediction module, wherein the first prediction module is used for inputting isomorphic characteristics of a target parking lot into a first layer model so as to output a description vector and a first prediction vector for describing flow trend, and the first layer model is a language sequence translation model; the input module is used for inputting the description vector, the first prediction vector and the heterogeneous characteristics of the target parking lot into a second-layer model, and obtaining a second prediction vector according to the description vector and the heterogeneous characteristics, wherein the second-layer model is a memory network model; and the second prediction module is used for inputting the first prediction vector into the second layer model and obtaining the predicted flow of the target parking lot based on the first prediction vector and the second prediction vector.

According to still another aspect of the present disclosure, there is provided a training apparatus for a traffic prediction model, including: the system comprises a first training module, a second training module and a third training module, wherein the first training module is used for performing model training of a language sequence translation model based on historical data of a plurality of parking lots to generate a first-layer model, the parking lots comprise target parking lots, and the historical data comprises isomorphic characteristics of the parking lots; the second training module is used for performing model training of a memory network on the output data of the first-layer model and the heterogeneous characteristics of the target parking lot in a transfer learning mode to obtain a second-layer model; a generating module for generating a traffic prediction model of the target parking lot based on the first layer model and the second layer model.

According to yet another aspect of the present disclosure, there is provided an electronic device including: a processor; and a memory for storing executable instructions for the processor; wherein the processor is configured to execute any one of the above-mentioned prediction methods of parking lot traffic and/or training methods of traffic prediction models via execution of executable instructions.

According to yet another aspect of the present disclosure, there is provided a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the method for predicting parking lot traffic and/or the method for training a traffic prediction model of any one of the above.

According to the parking lot traffic prediction scheme provided by the embodiment of the disclosure, a traffic prediction model comprising a first layer model and a second layer model is set, so that a description vector and a first prediction vector are obtained through a language sequence translation model in the first layer model, the description vector is used for reflecting the traffic trend of a target parking lot to be predicted, the first prediction vector is used for expressing a preliminary prediction result obtained based on isomorphic characteristics, further, the description vector and the first prediction vector are input into the second layer model, so that the description of the traffic trend based on the type of the target parking lot is obtained through the obtained specific heterogeneous characteristics of the target parking lot, and then a second prediction vector is obtained according to the description, and the second prediction vector is used for expressing a further prediction result based on the heterogeneous characteristics.

Furthermore, the predicted traffic of the target parking lot is obtained according to the first prediction vector and the second prediction vector, isomorphic cooperation processing of isomorphic features and heterogeneous features is achieved, and traffic prediction of the target parking lot is conducted, namely on the basis of prediction based on general performance of the parking lot, the heterogeneous features describing special performance of the parking lot are further added to serve as input parameters, traffic prediction is conducted on the basis of considering the type of the target parking lot, and accuracy of traffic prediction of the parking lot can be improved.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and together with the description, serve to explain the principles of the disclosure. It is to be understood that the drawings in the following description are merely exemplary of the disclosure, and that other drawings may be derived from those drawings by one of ordinary skill in the art without the exercise of inventive faculty.

Fig. 1 is a schematic diagram illustrating a prediction system structure of parking lot traffic in an embodiment of the present disclosure;

fig. 2 shows a flow chart of a method for predicting parking lot traffic in an embodiment of the present disclosure;

fig. 3 shows a schematic structural diagram of a second-layer model in the parking lot traffic prediction scheme according to the embodiment of the present disclosure;

fig. 4 shows a flow chart of another method for predicting parking lot traffic in an embodiment of the present disclosure;

fig. 5 shows a flowchart of a method for predicting parking lot traffic in another embodiment of the present disclosure;

fig. 6 shows a schematic structural diagram of a prediction model of parking lot traffic according to an embodiment of the present disclosure;

fig. 7 shows a flowchart of a method for predicting parking lot traffic in an embodiment of the present disclosure;

fig. 8 shows a flowchart of a method for predicting parking lot traffic in an embodiment of the present disclosure;

fig. 9 shows a schematic structural diagram of a first-layer model in a parking lot traffic prediction scheme according to an embodiment of the present disclosure;

FIG. 10 is a flow chart illustrating a method of training a traffic prediction model in an embodiment of the present disclosure;

FIG. 11 illustrates a training diagram of a traffic prediction model in an embodiment of the present disclosure;

FIG. 12 is a flow chart illustrating another method of training a traffic prediction model in an embodiment of the present disclosure;

FIG. 13 is a flow chart illustrating a method of training a flow prediction model according to yet another embodiment of the present disclosure;

FIG. 14 is a flow chart illustrating a method of training a flow prediction model according to yet another embodiment of the present disclosure;

FIG. 15 is a flow chart illustrating a method of training a flow prediction model according to yet another embodiment of the present disclosure;

fig. 16 is a schematic diagram illustrating a device for predicting parking lot traffic in an embodiment of the present disclosure;

FIG. 17 is a schematic diagram illustrating an apparatus for training a flow prediction model according to an embodiment of the disclosure;

fig. 18 shows a schematic diagram of an electronic device in an embodiment of the disclosure.

Detailed Description

Example embodiments will now be described more fully with reference to the accompanying drawings. Example embodiments may, however, be embodied in many different forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

Furthermore, the drawings are merely schematic illustrations of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus their repetitive description will be omitted. Some of the block diagrams shown in the figures are functional entities and do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in the form of software, or in one or more hardware modules or integrated circuits, or in different networks and/or processor devices and/or microcontroller devices.

According to the scheme, the forecasting flow of the target parking lot is obtained according to the first forecasting vector and the second forecasting vector, isomorphic cooperation processing of isomorphic characteristics and heterogeneous characteristics is achieved, flow forecasting of the target parking lot is conducted, namely on the basis of forecasting of general performance based on the parking lot, the heterogeneous characteristics describing special performance of the parking lot are further added to serve as input parameters, forecasting of traffic flow is conducted on the basis of considering the type of the target parking lot, and accuracy of traffic flow forecasting of the parking lot can be improved.

For ease of understanding, the following first explains several terms referred to in this application.

Artificial Neural Networks (ans), also referred to as Neural Networks (NNs) or Connection models (Connection models), are algorithmic mathematical models that Model animal Neural network behavior characteristics and perform distributed parallel information processing. The network achieves the aim of processing information by adjusting the mutual connection relationship among a large number of nodes in the network depending on the complexity of the system.

Machine Learning (ML) is a multi-domain cross discipline, and relates to a plurality of disciplines such as probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and the like. The special research on how a computer simulates or realizes the learning behavior of human beings so as to acquire new knowledge or skills and reorganize the existing knowledge structure to continuously improve the performance of the computer. Machine learning is the core of artificial intelligence, is the fundamental approach for computers to have intelligence, and is applied to all fields of artificial intelligence. Machine learning and deep learning generally include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and formal education learning.

The scheme provided by the embodiment of the application relates to technologies based on neural network modeling, machine learning and the like, and is specifically explained by the following embodiment.

Fig. 1 shows a schematic structural diagram of a parking lot traffic system in an embodiment of the present disclosure, which includes a plurality of terminals 120 and a server cluster 140.

The terminal 120 may be a mobile terminal such as a mobile phone, a game console, a tablet Computer, an e-book reader, smart glasses, an MP4(Moving Picture Experts Group Audio Layer IV) player, an intelligent home device, an AR (Augmented Reality) device, a VR (Virtual Reality) device, or a Personal Computer (PC), such as a laptop Computer and a desktop Computer.

Among them, an application program for providing parking lot traffic may be installed in the terminal 120.

The terminals 120 are connected to the server cluster 140 through a communication network. Optionally, the communication network is a wired network or a wireless network.

The server cluster 140 is a server, or is composed of a plurality of servers, or is a virtualization platform, or is a cloud computing service center. The server cluster 140 is used for providing background services for a prediction application program providing parking lot traffic and a training application program providing a traffic prediction model. Optionally, the server cluster 140 undertakes primary computational work and the terminal 120 undertakes secondary computational work; alternatively, the server cluster 140 undertakes secondary computing work and the terminal 120 undertakes primary computing work; alternatively, the terminal 120 and the server cluster 140 perform cooperative computing by using a distributed computing architecture.

In some optional embodiments, the server cluster 140 is used to store a prediction model and a prediction method of parking lot traffic, and the like.

Alternatively, the clients of the applications installed in different terminals 120 are the same, or the clients of the applications installed on two terminals 120 are clients of the same type of application of different control system platforms. Based on different terminal platforms, the specific form of the client of the application program may also be different, for example, the client of the application program may be a mobile phone client, a PC client, or a World Wide Web (Web) client.

Those skilled in the art will appreciate that the number of terminals 120 described above may be greater or fewer. For example, the number of the terminals may be only one, or several tens or hundreds of the terminals, or more. The number of terminals and the type of the device are not limited in the embodiments of the present application.

Optionally, the system may further include a management device (not shown in fig. 1), and the management device is connected to the server cluster 140 through a communication network. Optionally, the communication network is a wired network or a wireless network.

Optionally, the wireless network or wired network described above uses standard communication techniques and/or protocols. The Network is typically the Internet, but may be any Network including, but not limited to, a Local Area Network (LAN), a Metropolitan Area Network (MAN), a Wide Area Network (WAN), a mobile, wireline or wireless Network, a private Network, or any combination of virtual private networks. In some embodiments, data exchanged over a network is represented using techniques and/or formats including Hypertext Mark-up Language (HTML), Extensible markup Language (XML), and the like. All or some of the links may also be encrypted using conventional encryption techniques such as Secure Socket Layer (SSL), Transport Layer Security (TLS), Virtual Private Network (VPN), Internet protocol Security (IPsec). In other embodiments, custom and/or dedicated data communication techniques may also be used in place of, or in addition to, the data communication techniques described above.

Hereinafter, the steps in the parking lot traffic prediction method and the traffic prediction model training method according to the present exemplary embodiment will be described in more detail with reference to the drawings and examples.

Fig. 2 shows a flowchart of a method for predicting parking lot traffic in an embodiment of the present disclosure. The method provided by the embodiment of the present disclosure may be performed by any electronic device with computing processing capability, for example, the terminal 120 and/or the server cluster 140 in fig. 1. In the following description, the terminal 120 is taken as an execution subject for illustration.

As shown in fig. 2, the terminal 120 performs a method for predicting parking lot traffic, including the following steps:

step S202, isomorphic characteristics of the target parking lot are input into a first-layer model to output a description vector and a first prediction vector for describing flow trend, wherein the first-layer model is a language sequence translation model.

The language sequence translation model is a model for translating one language sequence into another language sequence, through language sequence translation, a description vector capable of describing the traffic trend of the parking lot can be obtained through isomorphic features extracted into the target parking lot, and a prediction result based on the language sequence translation model is a first prediction vector.

Isomorphic characteristics include, but are not limited to, time characteristics of different periods of time, flow characteristics of the period of time, parking space saturation of a parking lot, and the like, the time characteristics include long-term characteristics and short-term characteristics, the flow characteristics include inflow flow and outflow flow of the parking lot, and the isomorphic characteristics may also include weather characteristics, and the like.

For example, the language sequence translation model is a Seq2Seq network model.

Step S204, inputting the description vector, the first prediction vector and the heterogeneous characteristics of the target parking lot into a second-layer model, and obtaining a second prediction vector according to the description vector and the heterogeneous characteristics, wherein the second-layer model is a memory network model.

The heterogeneous characteristics include but are not limited to type characteristics of the parking lot, the type characteristics may include location types, business types and the like, taking the type characteristics as an example, the second layer model is a prediction model obtained by model training based on the memory network model, the memory network model is trained by adopting the type characteristics of the parking lots of different types to obtain the second layer model, description of traffic flow states of the parking lot of the type is obtained by combining description vectors output by the first layer model, and then corresponding second prediction vectors are obtained.

And step S206, obtaining the predicted flow of the target parking lot based on the first prediction vector and the second prediction vector.

In the embodiment, a description vector and a first prediction vector are obtained through a language sequence translation model in the first layer model by setting a flow prediction model comprising a first layer model and a second layer model, the description vector is used for reflecting the flow trend of a target parking lot to be predicted, the first prediction vector is used for expressing a preliminary prediction result obtained based on isomorphic features, further, the description vector and the first prediction vector are input into the second layer model to obtain a description of the vehicle flow trend based on the type of the target parking lot through the acquired specific heterogeneous features of the target parking lot, and then a second prediction vector is obtained according to the description, and the second prediction vector is used for expressing a further prediction result based on the heterogeneous features.

As shown in FIG. 3, in one embodiment, the memory network model includes an attention mechanism structure and a memory matrix, wherein the memory matrix is constructed based on different heterogeneous characteristics.

The Memory network model mainly comprises an attribute structure, namely an attention mechanism structure 302, and a Memory Matrix, namely a Memory Matrix 304, wherein the dimension of the Memory Matrix is a hyper-parameter, and the larger the parameter is, the higher the complexity of a recordable mode is.

The memory matrix can be understood as being composed of a plurality of vectors describing the flow trend, and each vector is used for describing the flow trend under one heterogeneous characteristic.

In addition, the memory network model can be replaced by a full-connection network, a neural cycle network RNN and other structures.

As shown in fig. 4, step 204, inputting the description vector, the first prediction vector, and the heterogeneous features of the target parking lot into the second-layer model, and obtaining a second prediction vector according to the description vector and the heterogeneous features includes:

in step S402, the description vector and the heterogeneous feature are input into the attention mechanism structure, and the matching operation is performed with the memory matrix.

As shown in fig. 3, the input of the layer is mainly the coding feature C generated by the Encoder in the first layer model, i.e. the description vector 306.

The heterogeneous characteristics 308 of the target parking lot include a location type characteristic and/or a business type characteristic of the target parking lot, and the like. Specifically, the attention mechanism structure attribute reads data in the Memory Matrix according to the correlation between the description vector and the heterogeneous features through the input description vector 306 and the heterogeneous features 308 to execute the matching operation.

Step S404, integrating the results of the matching operation to generate an integrated vector, where the integrated vector describes flow trends under different heterogeneous characteristics.

And matching the description vector with the matrix to obtain the trend of the traffic flow of the parking lot with the heterogeneous characteristics, and representing the trend in the form of an integrated vector.

Step S406, a second prediction vector is obtained based on the integration vector and the heterogeneous characteristics of the target parking lot.

As shown in fig. 3, the second prediction vector output by the attention mechanism structure is output by the output module 310, and can have the greatest correlation with the input heterogeneous features, that is, the prediction result based on the second prediction vector can have higher prediction accuracy.

In this embodiment, by inputting both the description vector and the heterogeneous features output by the first layer model into the second layer model, the output second prediction vector can take into account both the influence of the homogeneous features and the heterogeneous features on the traffic prediction, and a prediction result with higher accuracy can be obtained compared with a mode of predicting by only considering the homogeneous features.

In one embodiment, the prediction method further comprises: and configuring heterogeneous characteristics of the target parking lot based on the area type and/or the state type of the area where the target parking lot is located, wherein the area type comprises at least one of a transportation junction area, a business district area, an office area and a residential area.

The business type includes the location of the parking lot, the scale of the parking lot, the form of the parking lot, the price policy of the parking lot, and other type parameters.

For example, the heterogeneous characteristics of the parking lot in the transportation junction area may include dimensional characteristics of arrival-departure schedules of trains, buses, airplanes, etc., which may have a non-negligible effect on the current parking lot traffic.

In addition, the traffic flow in the parking lot is affected by characteristics such as obvious tidal property of the parking lot in the office area, shorter average parking time of the parking lot in the transportation hub area and the like.

In this embodiment, by introducing the above-mentioned heterogeneous features and specifically defining the heterogeneous features, the description of the flow prediction model is more accurate in combination with the isomorphic features in the first layer model.

As shown in fig. 5, in an embodiment, the traffic prediction model further adds a third layer model on the basis of the first layer model and the second layer model to perform a traffic prediction operation based on the third layer model, and the step S206 of obtaining the predicted traffic of the target parking lot based on the first prediction vector and the second prediction vector includes:

step S502, the first prediction vector and the second prediction vector are fused to obtain a fusion vector.

The method comprises the steps that a first prediction vector and a second prediction vector are obtained, the influence of isomorphic characteristics and heterogeneous characteristics of a parking lot on a prediction result can be fully considered, and the isomorphic characteristics and the heterogeneous characteristics are cooperatively predicted through fusion operation.

Step S504, inputting the fusion vector and the personality characteristics of the target parking lot into a third-layer model to execute full connection processing on the fusion vector and the personality characteristics, wherein the third-layer model is a full connection network model.

The parking lots with the same heterogeneous characteristics still have individual behaviors different from other parking lots, namely personalized characteristics, due to the fact that factors such as positions, environments and users are different, full connection processing is conducted on the fusion vector and the personalized characteristics, the full connection processing is equivalent to feature weighting conducted on the homogeneous characteristics, the heterogeneous characteristics and the personalized characteristics, a weighted characteristic is obtained, and due to the fact that the homogeneous characteristics, the heterogeneous characteristics and the personalized characteristics are taken into consideration, the prediction accuracy of flow prediction can be further improved.

Step S506 is to generate a predicted flow rate based on the processing result of the full link processing, and to cause the third layer model to output the predicted flow rate.

As shown in fig. 6, the traffic prediction model includes a first layer model 20, a second layer model 30 and a third layer model 30, in the first layer model 20, an encoder 602 outputs a description vector 306, the description vector 306 is input to a decoder 604 to obtain a first prediction vector, the description vector 306 is input to an attention mechanism 302 in the second layer model 30, a memory matrix 304 is combined to obtain a second prediction vector, and the first prediction vector and the second prediction vector are output to a fully-connected module 606 of the third layer model 40 through an output module 310 to obtain an output prediction result.

As shown in fig. 7, in an embodiment, the traffic prediction model may include only a first layer model and a second layer model, so as to perform a traffic prediction operation based on the first layer model and the second layer model, and the step S206 of obtaining the predicted traffic of the target parking lot based on the first prediction vector and the second prediction vector includes:

in step S702, a full concatenation process is performed on the first prediction vector and the second prediction vector.

In step S704, a predicted flow rate is generated based on the processing result of the all-connection processing, and the second layer model is caused to output the predicted flow rate.

In this embodiment, the flow prediction with higher accuracy is performed in time without training the third layer model by performing full-concatenation processing on the first prediction vector and the second prediction vector to obtain a processing value describing the predicted flow.

As shown in fig. 8, in an embodiment, the language sequence translation model includes a Seq2Seq network model, and the isomorphic characteristics of the target parking lot are input into the first layer model to output a description vector describing a traffic trend and a first prediction vector, and the first layer model is a language sequence translation model including:

step S802, isomorphic characteristics comprise time characteristics and corresponding flow characteristics, and the time characteristics and the flow characteristics are preprocessed to generate corresponding preprocessing vectors.

The seq2seq belongs to one of the language sequence translation models and comprises an encoder-decoder structure, wherein the encoder-decoder structure is an encoder-decoder structure, and a recurrent neural network model (RNN) is respectively arranged in the encoder and the decoder.

In addition, the coding structure and the decoding structure in the Encoder and the Decode can also use a long-time memory network LSTM network structure, a gated cyclic unit GRU network structure, a convolutional recurrent neural network CRNN network structure and the like to replace an RNN network structure.

Specifically, as shown in fig. 9, parking lot traffic and time characteristics 902 include holidays, workdays, weeks, hours, seasons, and the like, and the corresponding traffic characteristics include: long-time inflow flow Long-term inflow flow, Long-time outflow flow Long-term outflow flow, Short-term inflow flow, Short-term outflow flow, and other isomorphic characteristics may also include Schedule information and Weather information Weather.

As shown in fig. 9, the temporal features and the traffic features are preprocessed through embedding/full connection 904, specifically, embedding is performed on discrete features, normalization FC (full connection) is performed on continuous features, and finally, splicing is performed to form corresponding preprocessed vectors to be input into an Encoder or a Decoder.

Step S804, inputting the preprocessed vector into an encoder of the Seq2Seq network for encoding to obtain a description vector.

The encoder 602encoder is responsible for compressing an input sequence into a vector of a specified length, the vector can be regarded as the semantic of the sequence, the process is called encoding, and the manner of obtaining the semantic vector includes: directly taking the last input hidden state as a semantic vector C, performing a transformation on the last hidden state to obtain the semantic vector C, and performing a transformation on all hidden states of an input sequence to obtain a semantic variable C, wherein the semantic variable C is a description vector used in the disclosure.

Step S806, inputting the description vector and the prediction characteristics related to the prediction time into a decoder in the Seq2Seq network for decoding, so as to obtain a first prediction vector.

Further, based on the prediction characteristics of the description vector and the prediction time vector, a decoding operation is performed to obtain a corresponding specified sequence, so as to generate a first prediction vector.

In this embodiment, the first-layer model obtained based on the Seq2Seq network model has reusability of the model, and a description vector describing a traffic flow trend of the target parking lot is obtained in combination with isomorphic features of the target parking lot.

In one embodiment, in step S804, the encoder includes a first encoder, a second encoder, and a third encoder, and the inputting the preprocessed vector into an encoder of the Seq2Seq network for encoding, and the obtaining the description vector includes:

the pre-processing vector is input into a first encoder to encode the time characteristics of a plurality of timestamps before the prediction time and the corresponding flow characteristics to generate a first sequence.

And inputting the preprocessing vector into a second encoder to encode the time characteristics and the corresponding flow characteristics at the same time every day in a plurality of days before the preset time so as to generate a second sequence.

And inputting the preprocessing vector into a third encoder to encode the time characteristic and the corresponding flow characteristic at the same time every week in a plurality of weeks before the preset time so as to generate a third sequence.

As shown in fig. 9, the Encoder 602Encoder mines valid information for parking lot traffic and time characteristics 902, encodes the information to form a description vector 306, and inputs the description vector 306 to the Decoder 604Decoder as an initial state of the Decoder.

The Encoder 602Encoder is composed of three encoders, i.e., a first Encoder 6022, a second Encoder 6024, and a third Encoder 6026, which extract a sequence of the predicted time at the similar time, i.e., a first sequence, a sequence adjacent to the same time of day, i.e., a second sequence, and a sequence adjacent to the same time of week number, i.e., a third sequence, and a future correlation characteristic, respectively.

Specifically, for example, if the predicted time is ten am of the present thursday, the first sequence is a vector formed by predicting a plurality of time instants, such as nine, eight, seven, and six am of the present thursday, and the corresponding flow characteristics.

The second sequence is a vector formed by ten points of the day before the current day, ten points of the day before the current day and the like and corresponding flow characteristics.

The third sequence is a vector formed by the upper forty points of the week, and the like and the corresponding flow characteristics.

And obtaining the description vector based on the first sequence, the second sequence and the third sequence.

In the embodiment, the time-varying traffic trend of the target parking lot can be accurately described based on the description vector formed by the first sequence, the second sequence and the third sequence, and the basic traffic prediction is realized based on the traffic trend.

In one embodiment, step S806, inputting the description vector and the prediction characteristics related to the prediction time into a decoder in the Seq2Seq network for decoding, so as to obtain a first prediction vector, further includes:

as shown in fig. 9, the temporal characteristics of the predicted time 908 are isomorphically processed by the fully connected 906 module with the predicted result at the previous time to determine as the predicted characteristics.

And carrying out cooperative decoding on the description vector and the prediction characteristics to obtain a first prediction vector.

In the embodiment, the description vector is adopted as the initial state of the Decoder, and meanwhile, the time characteristic of the prediction time interval and the prediction result of the previous time are used as the input of the next prediction time and the input of the current prediction time interval into the Decoder, so that the supplementation and the enhancement of the Decoder input information are realized.

The existing parking lot flow prediction scheme can realize flow prediction of a certain parking lot or a certain type of parking lots, but the following problems still exist due to the fact that heterogeneous information among the parking lots is not fully considered:

(1) different parking lot data have different distributions and laws, and when flow prediction is carried out on different parking lots, the flow prediction needs to be carried out by combining with expert experience and the parking lot data of the parking lots, so that the reusability of the data and the model is not strong.

(2) And only a small amount of historical data or even no historical data exists in part of the parking lot, so that the existing parking lot flow prediction scheme cannot timely perform model training on such small sample scenes.

Based on the technical problem, the present disclosure also provides a method for training a traffic prediction model, so as to train the traffic prediction model in the above embodiment.

As shown in fig. 10, the terminal 120 executes a method for training a traffic prediction model, which includes the following steps:

step S1002, performing model training of the language sequence translation model based on historical data of a plurality of parking lots to generate a first-layer model, wherein the parking lots comprise target parking lots, and the historical data comprises isomorphic characteristics of the parking lots.

And step S1004, performing model training of a memory network on the output data of the first-layer model and the heterogeneous characteristics of the target parking lot in a transfer learning mode to obtain a second-layer model.

The output data of the first-layer model and the heterogeneous characteristics of the target parking lot are input into the memory network model, so that the first-layer model and the first-layer model are trained in a coordinated mode under the condition that the parameters of the first-layer model are fixed or the small learning rate is fine-tuned, and the first-layer model is trained by using transfer learning. In a target state, data used by Transfer learning is a data set with a large data volume and good quality as training data.

Step S1006, a flow prediction model of the target parking lot is generated based on the first layer model and the second layer model.

According to the training scheme of the flow prediction model, model training of the language order translation model is carried out through the historical data of the parking lot, a universal first-layer model in the flow prediction model is built, and therefore reusability of the first-layer model can be improved.

Further, model training of a memory network is performed based on the first-layer model and heterogeneous characteristics of the target parking lot, coexistence of common characteristics and characteristic characteristics of the parking lot in the traffic prediction model is achieved, accuracy of the obtained prediction model for parking lot traffic prediction is improved, the common characteristics comprise time characteristics, corresponding traffic characteristics and the like, and the characteristic characteristics can be heterogeneous characteristics.

As shown in fig. 12, in an embodiment, performing model training of the second-layer model by using a transfer learning method, specifically, the language sequence translation model includes a Seq2Seq network, and performing model training of the language sequence translation model based on historical data of a plurality of parking lots in step S1002, and generating the first-layer model by using the traffic prediction model includes:

and step S1202, performing model training of the Seq2Seq network through the second isomorphic features to obtain a basic model, wherein the isomorphic features of the parking lot comprise the first isomorphic feature of the target parking lot and the second isomorphic features of other parking lots.

Step S1204, transfer the training result of the basic model to the first layer model to be trained;

and step S1206, performing model training on the first layer model to be trained through the first isomorphic characteristic to obtain the trained first layer model.

In this embodiment, the first-level model is trained in a transfer learning manner, so that reusability of the model and training data can be ensured, and the training data is reused to obtain the first-level model with better generalization performance when the target parking lot is favored by a small-sample parking lot.

As shown in fig. 13, in an embodiment, performing model training of the second layer model in a meta-learning manner, wherein the language sequence translation model includes a Seq2Seq network, and performing model training of the language sequence translation model based on historical data of a plurality of parking lots to generate the first layer model of the traffic prediction model includes:

step S1302, the isomorphic characteristics of the parking lot comprise a first isomorphic characteristic of the target parking lot and second isomorphic characteristics of other parking lots, and a plurality of training task sets are generated according to the second isomorphic characteristics.

In step S1304, a meta-learning model of the Seq2Seq network is generated from the plurality of training task sets.

Step 1306, model learning is conducted on the meta-learning model based on the first isomorphic feature, and a first layer model is obtained.

In this embodiment, the data of the used parking lot can be used for training the meta-learning model, and the learning mode does not need to carry out excessive screening on the historical data of the parking lot and has less limitation on the data quantity. The basic model of the layer is trained by using the meta-learning model, and all the state parking lots can share one basic model. The training data may be trained using all available data. After training is finished, the method can be applied to new parking lot prediction business by adjusting for several rounds according to the state of business or a certain parking lot. Compared with the transfer learning mode, the meta learning method has stronger generalization, lower requirements on the quality and quantity of training data and less consumed resources.

As shown in fig. 14, in an embodiment, in step S1004, performing model training of a memory network on output data of the first-layer model and heterogeneous features of the target parking lot in a migration learning manner, and obtaining the second-layer model includes:

step S1402, performing model training on the memory network according to the training parameters of the first layer model to obtain the structural parameters of the attention structure and the memory matrix in the memory network.

Step S1404, the structural parameters and the memory matrix are transferred to a second-layer model to be trained, and model training is carried out on the second-layer model to be trained through heterogeneous characteristics of the target parking lot, so that the trained second-layer model is obtained.

In the embodiment, the parameters needing to be trained in the layer are mainly an attribute structure and a Memory Matrix, and the parameters are isomorphically added into a Memory module to learn the heterogeneous characteristics of different parking lots, so that the contradiction between a general model and heterogeneous data can be solved.

As shown in fig. 15, in one embodiment, the step S1006 of generating a traffic prediction model of the target parking lot based on the first layer model and the second layer model includes:

step S1502, model training of a full-connection network is conducted on output data of the first-layer model, output data of the second-layer model and personality characteristics of the target parking lot, and a third-layer model is obtained.

Step S1504, a traffic prediction model is generated based on the first layer model, the second layer model, and the third layer model.

As shown in fig. 11, the first-level model 20 is trained using a plurality of pieces of parking lot data, that is, all pieces of parking lot data 1102 and external features 1104, the second-level model 30 is trained using the output of the first-level model 20 and one piece of parking lot data 1106 and external features 1108, and the third-level model 40 is trained using the output of the second-level model 30, one piece of parking lot data 1110 and one piece of external feature data 1112, and the flow prediction model 50 is obtained after the training of the flow prediction model is completed.

In this embodiment, the present layer model mainly aims at individual training, and each parking lot differs depending on the location, environment, and user. Therefore, the current layer mainly excavates the behavior mode of the current parking lot and fuses the outputs of the models of the previous two layers. The layer mainly comprises a plurality of full connection layers.

Specifically, the input of the current layer is mainly the output data of the first two layers of models and the characteristic features of the target parking. The output of the current layer is the predicted flow of the target parking lot.

And the training of the model of the third layer is performed under the condition that the training is performed simultaneously in cooperation with the training of the two previous layers and the parameters of the two previous layers are fixed or the fine adjustment of the small learning rate is performed. The modules of the training of the layer are mainly fully connected. The used data is historical behavior data of the target parking lot.

In addition, the model uses the mean square loss as a loss function of training, the true value is a true sequence of future flow at the current node, and the calculation formula is expressed as:

where m is the number of samples and n is the predicted sequence length. Y is_ijIn order to be a vector of true values,

is a predictor vector.

The model can be used for predicting the flow of different parking lots, particularly reconstructed parking lots and newly added parking lots, the reusability of data and the model is improved while the prediction accuracy is ensured, and the customization cost is reduced. The method can be further used for calculating a parking lot income prediction curve, a parking space saturation prediction curve, a parking index and the like, provides predicted parking information for a driver to improve travel convenience, and provides future parking passenger flow change and parking space resource information for a manager to reduce operation difficulty.

It is to be noted that the above-mentioned figures are only schematic illustrations of the processes involved in the method according to an exemplary embodiment of the invention, and are not intended to be limiting. It will be readily understood that the processes shown in the above figures are not intended to indicate or limit the chronological order of the processes. In addition, it is also readily understood that these processes may be performed synchronously or asynchronously, e.g., in multiple modules.

As will be appreciated by one skilled in the art, aspects of the present invention may be embodied as a system, method or program product. Thus, various aspects of the invention may be embodied in the form of: an entirely hardware embodiment, an entirely software embodiment (including firmware, microcode, etc.) or an embodiment combining hardware and software aspects that may all generally be referred to herein as a "circuit," module "or" system.

A prediction apparatus 1600 of parking lot traffic according to this embodiment of the present invention is described below with reference to fig. 16. The parking lot traffic device 1600 shown in fig. 16 is only an example, and should not bring any limitation to the function and the range of use of the embodiment of the present invention.

The parking lot traffic prediction device 1600 is represented in the form of a hardware module. The components of the prediction device 1600 of parking lot traffic may include, but are not limited to: a first prediction module 1602, configured to input isomorphic features of the target parking lot into a first layer model to output a description vector describing a traffic trend and a first prediction vector, where the first layer model is a language sequence translation model; the input module 1604 is configured to input the description vector, the first prediction vector, and the heterogeneous features of the target parking lot into a second-layer model, and obtain a second prediction vector according to the description vector and the heterogeneous features, where the second-layer model is a memory network model; and a second prediction module 1606, configured to obtain the predicted traffic of the target parking lot based on the first prediction vector and the second prediction vector.

In one embodiment, the input module 1604 is further for: inputting the description vector and the heterogeneous characteristics into an attention mechanism structure, and performing matching operation with a memory matrix; integrating the results of the matching operation to generate an integrated vector, wherein the integrated vector describes the flow trend under different heterogeneous characteristics; and obtaining a second prediction vector based on the integration vector and the heterogeneous characteristics of the target parking lot.

In one embodiment, the device 1600 for predicting parking lot traffic further comprises: a configuration module 1608 configured to configure heterogeneous characteristics of the target parking lot based on an area type and/or an enterprise type of an area in which the target parking lot is located, wherein the area type includes at least one of a transportation hub area, a business district area, an office district area, and a residential district area.

In one embodiment, the second prediction module 1606 is further configured to: performing fusion processing on the first prediction vector and the second prediction vector to obtain a fusion vector; the prediction apparatus 1600 for parking lot traffic further includes: a third prediction module 1610, configured to input the fusion vector and the personality characteristics of the target parking lot into a third-layer model, so as to perform full connection processing on the fusion vector and the personality characteristics, where the third-layer model is a full connection network model; and generating a predicted flow based on the processing result of the full-connection processing, and enabling the third-layer model to output the predicted flow.

In one embodiment, the second prediction module 1606 is further configured to: fully connecting the first prediction vector and the second prediction vector; and generating a predicted flow rate based on the processing result of the full-connection processing, and enabling the second layer model to output the predicted flow rate.

In one embodiment, the first prediction module 1602 is further configured to: the isomorphic characteristics comprise time characteristics and corresponding flow characteristics, and the time characteristics and the flow characteristics are preprocessed to generate corresponding preprocessed vectors; inputting the preprocessed vector into an encoder of a Seq2Seq network for encoding to obtain a description vector; and inputting the description vector and the prediction characteristics related to the prediction time into a decoder in the Seq2Seq network for decoding to obtain a first prediction vector.

In one embodiment, the first prediction module 1602 is further configured to: inputting the preprocessing vector into a first encoder to encode time characteristics of a plurality of timestamps before the prediction time and corresponding flow characteristics to generate a first sequence; inputting the preprocessing vector into a second encoder to encode time characteristics and corresponding flow characteristics at the same time every day in multiple days before preset time to generate a second sequence; inputting the preprocessed vector into a third encoder to encode the time characteristic and the corresponding flow characteristic at the same time every week in a plurality of weeks before the preset time so as to generate a third sequence; and obtaining the description vector based on the first sequence, the second sequence and the third sequence.

In one embodiment, the first prediction module 1602 is further configured to: determining the time characteristic of the predicted time and the predicted result of the previous moment as predicted characteristics; and carrying out cooperative decoding on the description vector and the prediction characteristics to obtain a first prediction vector.

The following describes the training apparatus 1700 of the flow prediction model according to this embodiment of the present invention with reference to fig. 17. The training apparatus 1700 of the flow prediction model shown in fig. 17 is only an example, and should not bring any limitation to the function and the scope of the embodiment of the present invention.

The training means 1700 of the flow prediction model is represented in the form of a hardware module. The components of the training apparatus 1700 of the flow prediction model may include, but are not limited to: a first training module 1702, configured to perform model training of the language order translation model based on historical data of a plurality of parking lots to generate a first-layer model, where the plurality of parking lots includes a target parking lot, and the historical data includes isomorphic features of the plurality of parking lots; a second training module 1704, configured to perform model training of a memory network on output data of the first-layer model and heterogeneous features of the target parking lot in a transfer learning manner, so as to obtain a second-layer model; a generating module 1706 is configured to generate a traffic prediction model of the target parking lot based on the first-layer model and the second-layer model.

In one embodiment, the language translation model includes a Seq2Seq network, and the first training module 1702 is further configured to: the historical data of the parking lots comprise first isomorphic characteristics of a target parking lot and second isomorphic characteristics of other parking lots, and model training of a Seq2Seq network is carried out through the second isomorphic characteristics to obtain a basic model; and transferring the training result of the basic model to a first layer model to be trained, and performing model training on the first layer model to be trained through the first isomorphic characteristic to obtain the trained first layer model.

In one embodiment, the language translation model includes a Seq2Seq network, and the first training module 1702 is further configured to: the historical data of the parking lots comprises a first isomorphic feature of the target parking lot and second isomorphic features of other parking lots, and a plurality of training task sets are generated according to the second isomorphic features; generating a meta-learning model of the Seq2Seq network according to a plurality of training task sets; and performing model learning based on the first isomorphic feature element learning model pair to obtain a first layer model.

In one embodiment, the second training module 1704 is further configured to: performing model training of the memory network through the training parameters of the first layer of model to obtain structural parameters and a memory matrix of the attention structure in the memory network; and migrating the structural parameters and the memory matrix to a second-layer model to be trained, and performing model training on the second-layer model to be trained through the heterogeneous characteristics of the target parking lot to obtain the trained second-layer model.

In one embodiment, the apparatus 1700 for training the flow prediction model further includes: a third training module 1708, configured to perform model training of a fully connected network on output data of the first-layer model, output data of the second-layer model, and personality characteristics of the target parking lot, to obtain a third-layer model; the generating module 1706 is further configured to: and generating a flow prediction model based on the first layer model, the second layer model and the third layer model.

An electronic device 1800 according to this embodiment of the invention is described below with reference to fig. 18. The electronic device 1800 shown in fig. 18 is only an example, and should not bring any limitations to the function and the scope of use of the embodiments of the present invention.

As shown in fig. 18, the electronic device 1800 is in the form of a general purpose computing device. Components of the electronic device 1800 may include, but are not limited to: the at least one processing unit 1810, the at least one memory unit 1820, and the bus 1830 that couples the various system components including the memory unit 1820 and the processing unit 1810.

Where the memory unit stores program code, the program code may be executed by the processing unit 1010 to cause the processing unit 1810 to perform steps according to various exemplary embodiments of the present invention described in the above section "exemplary method" of the present specification. For example, the processing unit 1010 may perform steps S202, S204, and S206 as shown in fig. 2, and other steps defined in the disclosed parking lot traffic prediction method and/or the traffic prediction model training method.

The storage unit 1820 may include readable media in the form of volatile storage units, such as a random access memory unit (RAM)18201 and/or a cache memory unit 18202, and may further include a read-only memory unit (ROM) 18203.

The storage unit 1820 may also include a program/utility 18204 having a set (at least one) of program modules 18205, such program modules 18205 including, but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which, or some combination thereof, may comprise an implementation of a network environment.

The bus 1830 may be any type of bus structure representing one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

The electronic device 1800 may also communicate with one or more external devices 1860 (e.g., keyboard, pointing device, bluetooth device, etc.), with one or more devices that enable a user to interact with the electronic device, and/or with any device (e.g., router, modem, etc.) that enables the electronic device 1800 to communicate with one or more other computing devices. Such communication can occur through input/output (I/O) interface 1850. Also, the electronic device 1800 may communicate with one or more networks (e.g., a Local Area Network (LAN), a Wide Area Network (WAN), and/or a public network such as the internet) via the network adapter 1850. As shown, the network adapter 1850 communicates with other modules of the electronic device 1800 via the bus 1830. It should be appreciated that although not shown in the figures, other hardware and/or software modules may be used in conjunction with the electronic device, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, among others.

Through the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein may be implemented by software, or by software in combination with necessary hardware. Therefore, the technical solution according to the embodiments of the present disclosure may be embodied in the form of a software product, which may be stored in a non-volatile storage medium (which may be a CD-ROM, a usb disk, a removable hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which may be a personal computer, a server, a terminal device, or a network device, etc.) to execute the method according to the embodiments of the present disclosure.

In an exemplary embodiment of the present disclosure, there is also provided a computer-readable storage medium having stored thereon a program product capable of implementing the above-described method of the present specification. In some possible embodiments, aspects of the invention may also be implemented in the form of a program product comprising program code means for causing a terminal device to carry out the steps according to various exemplary embodiments of the invention described in the above-mentioned "exemplary methods" section of the present description, when the program product is run on the terminal device.

According to the program product for realizing the method, the portable compact disc read only memory (CD-ROM) can be adopted, the program code is included, and the program product can be operated on terminal equipment, such as a personal computer. However, the program product of the present invention is not limited in this regard and, in the present document, a readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.

A computer readable signal medium may include a propagated data signal with readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A readable signal medium may also be any readable medium that is not a readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.

Program code embodied on a readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

Program code for carrying out operations for aspects of the present invention may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device and partly on a remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computing device (e.g., through the internet using an internet service provider).

It should be noted that although in the above detailed description several modules or units of the device for action execution are mentioned, such a division is not mandatory. Indeed, the features and functionality of two or more modules or units described above may be embodied in one module or unit, according to embodiments of the present disclosure. Conversely, the features and functions of one module or unit described above may be further divided into embodiments by a plurality of modules or units.

Moreover, although the steps of the methods of the present disclosure are depicted in the drawings in a particular order, this does not require or imply that the steps must be performed in this particular order, or that all of the depicted steps must be performed, to achieve desirable results. Additionally or alternatively, certain steps may be omitted, multiple steps combined into one step execution, and/or one step broken down into multiple step executions, etc.

Through the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein may be implemented by software, or by software in combination with necessary hardware. Therefore, the technical solution according to the embodiments of the present disclosure may be embodied in the form of a software product, which may be stored in a non-volatile storage medium (which may be a CD-ROM, a usb disk, a removable hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which may be a personal computer, a server, a mobile terminal, or a network device, etc.) to execute the method according to the embodiments of the present disclosure.

Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This application is intended to cover any variations, uses, or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.

Claims

1. A method for predicting parking lot traffic is characterized by comprising the following steps:

inputting isomorphic characteristics of a target parking lot into a first-layer model to output a description vector and a first prediction vector for describing a flow trend, wherein the first-layer model is a language sequence translation model;

inputting the description vector, the first prediction vector and the heterogeneous features of the target parking lot into a second-layer model, and obtaining a second prediction vector according to the description vector and the heterogeneous features; and

and obtaining the predicted flow of the target parking lot based on the second prediction vector and the first prediction vector, wherein the second layer model is a memory network model.

2. The parking lot traffic prediction method according to claim 1, wherein the memory network model includes an attention mechanism structure and a memory matrix, wherein the memory matrix is formed based on different heterogeneous characteristics;

the obtaining a second prediction vector according to the description vector and the heterogeneous features comprises:

inputting the description vector and the heterogeneous features into the attention mechanism structure, and performing matching operation with the memory matrix;

integrating results of the matching operations to generate an integrated vector, wherein the integrated vector describes the flow trend under the different heterogeneous characteristics;

and obtaining the second prediction vector based on the integration vector and the heterogeneous characteristics of the target parking lot.

3. The method for predicting parking lot traffic according to claim 1, further comprising:

configuring heterogeneous features of the target parking lot based on the area type and/or the business state type of the area where the target parking lot is located,

wherein the area type includes at least one of a transportation junction area, a business district area, an office area, and a residential area.

4. The method for predicting traffic of parking lot according to claim 1, wherein the obtaining the predicted traffic of the target parking lot based on the first prediction vector and the second prediction vector comprises:

performing fusion processing on the first prediction vector and the second prediction vector to obtain a fusion vector;

inputting the fusion vector and the personality characteristics of the target parking lot into a third-layer model to perform full-connection processing on the fusion vector and the personality characteristics, wherein the third-layer model is a full-connection network model;

and generating the predicted flow based on the processing result of the full-connection processing, and enabling the third-layer model to output the predicted flow.

5. The method for predicting traffic of parking lot according to claim 1, wherein the obtaining the predicted traffic of the target parking lot based on the first prediction vector and the second prediction vector comprises:

performing full-concatenation processing on the first prediction vector and the second prediction vector;

and generating the predicted flow based on the processing result of the full-connection processing, and enabling the second layer model to output the predicted flow.

6. The method according to any one of claims 1 to 5, wherein the language translation model comprises a Seq2Seq network;

the inputting isomorphic features of the target parking lot into the first layer model to output a description vector describing the traffic trend and a first prediction vector comprises:

the isomorphic features comprise time features and corresponding flow features, and the time features and the flow features are preprocessed to generate corresponding preprocessing vectors;

inputting the preprocessed vector into an encoder of the Seq2Seq network for encoding to obtain the description vector;

and inputting the description vector and the prediction characteristics related to the prediction time into a decoder in the Seq2Seq network for decoding to obtain the first prediction vector.

7. The method for predicting parking lot traffic according to claim 6, wherein the encoder includes a first encoder, a second encoder, and a third encoder;

inputting the preprocessed vector into an encoder of the Seq2Seq network for encoding to obtain the description vector comprises:

inputting the preprocessing vector into the first encoder to encode time characteristics of a plurality of timestamps before the prediction time and corresponding flow characteristics to generate a first sequence;

inputting the preprocessing vector into the second encoder to encode the time characteristics and the corresponding flow characteristics at the same time every day in a plurality of days before the preset time so as to generate a second sequence;

inputting the preprocessing vector into the third encoder to encode the time characteristics and the corresponding flow characteristics at the same time every week in a plurality of weeks before the preset time so as to generate a third sequence;

obtaining the description vector based on the first sequence, the second sequence and the third sequence.

8. The method for predicting parking lot traffic according to claim 6, wherein the step of inputting the description vector and the prediction feature related to the prediction time into a decoder in the Seq2Seq network to decode the description vector and the prediction feature to obtain the first prediction vector comprises:

determining the time characteristic of the prediction time and the prediction result of the previous moment as the prediction characteristic;

and carrying out cooperative decoding on the description vector and the prediction characteristics to obtain the first prediction vector.

9. A method for training a flow prediction model is characterized by comprising the following steps:

performing model training of a language sequence translation model based on historical data of a plurality of parking lots to generate a first-layer model, wherein the parking lots comprise target parking lots, and the historical data comprises isomorphic features of the parking lots;

performing model training of a memory network on the output data of the first-layer model and the type characteristics of the target parking lot in a transfer learning mode to obtain a second-layer model;

and generating a flow prediction model of the target parking lot based on the first layer model and the second layer model.

10. The method for training a traffic prediction model according to claim 9, wherein the language sequence translation model comprises a Seq2Seq network;

the model training of the language sequence translation model based on the historical data of the plurality of parking lots, and the generation of the first-layer model by the flow prediction model comprises the following steps:

the historical data of the parking lots comprise first isomorphic characteristics of the target parking lot and second isomorphic characteristics of other parking lots, and model training of the Seq2Seq network is carried out through the second isomorphic characteristics to obtain a basic model;

and transferring the training result of the basic model to the first layer model to be trained, and performing model training on the first layer model to be trained through the first isomorphic feature to obtain the trained first layer model.

11. The method for training a traffic prediction model according to claim 9, wherein the language sequence translation model comprises a Seq2Seq network;

the model training of the language sequence translation model based on the historical data of the plurality of parking lots to generate the first-layer model of the flow prediction model comprises:

the historical data of the parking lots comprises a first isomorphic feature of the target parking lot and a second isomorphic feature of other parking lots, and a plurality of training task sets are generated according to the second isomorphic feature;

generating a meta-learning model of the Seq2Seq network according to the plurality of training task sets;

and carrying out model learning on the meta-learning model pair based on the first isomorphic characteristic to obtain the first layer model.

12. The method for training the traffic prediction model according to claim 9, wherein the performing model training of the memory network on the output data of the first-layer model and the heterogeneous characteristics of the target parking lot in the transfer learning manner to obtain the second-layer model comprises:

performing model training on the memory network through the training parameters of the first layer model to obtain structural parameters and a memory matrix of an attention structure in the memory network;

and migrating the structural parameters and the memory matrix to the second-layer model to be trained, and performing model training on the second-layer model to be trained through the heterogeneous characteristics of the target parking lot to obtain the trained second-layer model.

13. The method for training the traffic prediction model according to any one of claims 9 to 12, wherein the generating the traffic prediction model of the target parking lot based on the first-layer model and the second-layer model includes:

performing model training of a fully-connected network on the output data of the first-layer model, the output data of the second-layer model and the individual characteristics of the target parking lot to obtain a third-layer model;

generating the traffic prediction model based on the first layer model, the second layer model, and the third layer model.

14. A prediction device of parking lot traffic, comprising:

the system comprises a first prediction module, a second prediction module and a third prediction module, wherein the first prediction module is used for inputting isomorphic characteristics of a target parking lot into a first layer model so as to output a description vector and a first prediction vector for describing flow trend, and the first layer model is a language sequence translation model;

the input module is used for inputting the description vector, the first prediction vector and the heterogeneous characteristics of the target parking lot into a second-layer model so as to obtain a second prediction vector according to the description vector and the heterogeneous characteristics;

and the second prediction module is used for obtaining the predicted flow of the target parking lot according to the second prediction vector and the first prediction vector, and the second layer model is a memory network model.

15. An apparatus for training a flow prediction model, comprising:

the system comprises a first training module, a second training module and a third training module, wherein the first training module is used for performing model training of a language sequence translation model based on historical data of a plurality of parking lots to generate a first-layer model, the parking lots comprise target parking lots, and the historical data comprises isomorphic characteristics of the parking lots;

the second training module is used for performing model training of a memory network on the output data of the first-layer model and the heterogeneous characteristics of the target parking lot in a transfer learning mode to obtain a second-layer model;

a generating module for generating a traffic prediction model of the target parking lot based on the first layer model and the second layer model.

16. An electronic device, comprising:

a processor; and

a memory for storing executable instructions of the processor;

wherein the processor is configured to execute the method for predicting parking lot traffic according to any one of claims 1 to 8 and/or the method for training traffic prediction model according to any one of claims 9 to 13 via executing the executable instructions.

17. A computer-readable storage medium, on which a computer program is stored, wherein the computer program, when executed by a processor, implements the method for predicting a traffic flow of a parking lot according to any one of claims 1 to 8 and/or the method for training a traffic prediction model according to any one of claims 9 to 13.