Scenarized supply and demand docking method for data elements

Abstract

The invention discloses a data element scenarized supply and demand docking method, and relates to the technical field of supply and demand matching in a data element market. The method comprises the following steps: collecting original data of a supplier and a demander, and carrying out denoising and formatting processing on the original data to obtain a preprocessed supplier data set and a preprocessed demander data set; constructing a scenarized data model based on the convolutional neural network and a deep auto-encoder, and obtaining a supplier feature vector and a demand feature vector after scenarized reconstruction by using the scenarized data model; and constructing a vector knowledge base to store the supplier feature vector and the demander feature vector after scene reconstruction, and obtaining a final supply-demand docking result by using a supply-demand docking algorithm based on the actual demand task and the vector knowledge base. According to the invention, the technical problems of low supply and demand matching efficiency, serious information island phenomenon and insufficient scenarized application capability in the current data element market are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of supply-demand matching in the data element market, and specifically relates to a method for scenario-based supply-demand docking of data elements. Background Art

[0002] In the current data element market, there are problems such as low supply-demand matching efficiency, serious information island phenomenon, and insufficient scenario application ability. Traditional data element markets often face challenges such as information asymmetry between supply and demand sides, difficulty in data value evaluation, and difficulty in quickly matching data application scenarios. This has led to waste of data resources and limited data-driven business innovation. Summary of the Invention

[0003] In view of the above deficiencies in the prior art, the present invention provides a method for scenario-based supply-demand docking of data elements, which solves the technical problems of low supply-demand matching efficiency, serious information island phenomenon, and insufficient scenario application ability existing in the current data element market.

[0004] In order to achieve the above invention purpose, the technical solution adopted by the present invention is as follows: A method for scenario-based supply-demand docking of data elements, comprising the following steps: S1. Collect the original data of the supply side and the demand side, perform denoising and formatting processing on the original data of the supply side and the demand side to obtain the preprocessed supply-side data set and demand-side data set; S2. Construct a scenario-based data model based on a convolutional neural network and a deep autoencoder, and use the scenario-based data model, the preprocessed supply-side data set, and the demand-side data set to obtain the supply-side feature vector and demand-side feature vector after scenario-based reconstruction; S3. Construct a vector knowledge base to store the supply-side feature vector and demand-side feature vector after scenario-based reconstruction, and based on the actual demand task and the vector knowledge base, use a supply-demand docking algorithm to obtain the final supply-demand docking result.

[0005] Further, in step S2, the scenario-based data model includes a first input layer, a convolutional layer, a pooling layer, a fully connected layer, a first output layer, a second input layer, an encoding layer, a latent space representation layer, a decoding layer, and a second output layer connected in sequence.

[0006] Further, the data processing process of the scenario-based data model is specifically as follows: The first input layer is used to receive the first input data, and the first input data is the preprocessed supply-side data; The convolutional layer is used to perform feature extraction on the first input data through a convolutional kernel to obtain a convolutional output, expressed as: C = σ (W c ·X1 + bc ) Where: C is the convolutional output of the supplier data, σ ( ) is the ReLU activation function, W c is the weight coefficient of the convolutional layer, X1 is the preprocessed supplier data, b c is the bias coefficient of the convolutional layer; The pooling layer is used to downsample the convolutional output of the supplier data through max-pooling operation to obtain the pooling feature of the supplier data, expressed as: P = MaxPool(C) Where: P is the pooling feature of the supplier data, MaxPool is the max-pooling operation; The fully connected layer is used to flatten the pooling feature of the supplier data and obtain the mapping feature of the supplier data through the weight coefficient, expressed as: F = σ (W f ·P + b f ) Where: F is the mapping feature of the supplier data, W f is the weight coefficient of the fully connected layer, b f is the bias coefficient of the fully connected layer; The first output layer is used to generate the supplier feature vector according to the mapping feature of the supplier data, expressed as: V s = W·F + b o Where: V s is the supplier feature vector, W is the weight coefficient of the first output layer, b o is the bias coefficient of the first output layer; The second input layer is used to receive the second input data, and the second input data is the preprocessed demander data; The encoding layer is used to perform non-linear compression on the second input data to obtain the encoded data of the demander data, expressed as: E = σ (W e ·X2 + b e ) Where: E is the encoded data of the demander data, W e is the weight coefficient of the encoding layer, X2 is the preprocessed demander data, b e is the bias coefficient of the encoding layer; The latent space representation layer is used to reduce the dimension and extract features from the encoded data of the demander data to generate the low-dimensional feature of the demander data, expressed as: Z = W z ·E + b z Where: Z is the low-dimensional feature of the demander data, and W z is the weight coefficient of the latent space representation layer, and b z is the bias coefficient of the latent space representation layer; The decoding layer is used to reconstruct the low-dimensional feature of the demander data to obtain the reconstructed data of the demander data, which is expressed as: D = σ (W d ·Z + b d ) Where: D is the reconstructed data of the demander data, and W d is the weight coefficient of the decoding layer, and b d is the bias coefficient of the decoding layer; The second output layer is used to generate an initial demander feature vector according to the reconstructed data of the demander data, which is expressed as: V d = W ′ d ·D + b ′ d Where: V d is the initial demander feature vector, and W ′ d is the weight coefficient of the second output layer, and b ′ d is the bias coefficient of the second output layer; And the initial demander feature vector is normalized through the Sigmoid function to obtain the demander feature vector.

[0007] Furthermore, in step S3, based on the actual demand task and the vector knowledge base, the final supply-demand matching result is obtained by using the supply-demand matching algorithm, including the following steps: A1. Based on the vector knowledge base, use the supply-demand matching algorithm to calculate the similarity between the demander and the supplier; A2. Set the similarity threshold between the demander and the supplier based on the actual demand task; A3. Based on the similarity between the demander and the supplier and the similarity threshold between the demander and the supplier, obtain the final supply-demand matching result.

[0008] Furthermore, in step A1, based on the vector knowledge base, use the supply-demand matching algorithm to calculate the similarity between the demander and the supplier, which is expressed as:

[0009] Where: is the similarity between the demander and the supplier, is the th scenario-reconstructed supplier feature vector, is the demand-side feature vector after the th scenario-based reconstruction, is the norm of the th scenario-based reconstructed supply-side feature vector, is the norm of the th scenario-based reconstructed demand-side feature vector.

[0010] Furthermore, in step S3, in the process of obtaining the final supply-demand matching result by using the supply-demand matching algorithm based on the actual demand task and the vector knowledge base, an encryption algorithm is used to encrypt the actual demand task and the final supply-demand matching result.

[0011] The present invention has the following beneficial effects: (1) By constructing a scenario-based data model according to the convolutional neural network and the deep autoencoder, and using the scenario-based data model, the preprocessed supply-side data set and demand-side data set, the present invention obtains the scenario-based reconstructed supply-side feature vector and demand-side feature vector, and solves the technical problem of insufficient scenario-based application ability in the current data element market; (2) By constructing a vector knowledge base to store the scenario-based reconstructed supply-side feature vector and demand-side feature vector, and based on the actual demand task and the vector knowledge base, using the supply-demand matching algorithm to obtain the final supply-demand matching result, the present invention solves the problems of low supply-demand matching efficiency and serious information island phenomenon in the current data element market. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 is a schematic flow chart of a data element scenario-based supply-demand matching method. DETAILED DESCRIPTION OF THE INVENTION

[0013] The following describes the specific embodiments of the present invention to facilitate those skilled in the art to understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention defined and determined by the appended claims, these changes are obvious, and all inventions made using the inventive concept of the present invention are within the scope of protection.

[0014] As Figure 1 shown, a data element scenario-based supply-demand matching method includes steps S1-S3, which are specifically as follows: S1. Collect the original data of the supply side and the demand side, and perform denoising and formatting processing on the original data of the supply side and the demand side to obtain the preprocessed supply-side data set and demand-side data set.

[0015] In an alternative embodiment of the present invention, the original data of the supplier and the demander comes from multiple different fields and industries, including finance, healthcare, education, and transportation. The data types include tabular data in structured data and XML-format data and JSON-format data in semi-structured data. In the original data of the supplier and the demander, there is redundant and invalid information, and there is also noise. The noise includes incorrect data records and incomplete data items. The present invention performs denoising and formatting processing on the original data of the supplier and the demander to obtain the preprocessed supplier dataset and demander dataset, so as to remove redundant, invalid information, and the interference of noise.

[0016] S2. Construct a scenario-based data model based on a convolutional neural network and a deep autoencoder, and use the scenario-based data model, the preprocessed supplier dataset, and the demander dataset to obtain the scenario-based reconstructed supplier feature vector and demander feature vector.

[0017] In an alternative embodiment of the present invention, the scenario-based data model includes a first input layer, a convolutional layer, a pooling layer, a fully connected layer, a first output layer, a second input layer, an encoding layer, a latent space representation layer, a decoding layer, and a second output layer connected in sequence.

[0018] The data processing process of the scenario-based data model is specifically as follows: The first input layer is used to receive the first input data, and the first input data is the preprocessed supplier data.

[0019] The convolutional layer is used to extract features from the first input data through a convolutional kernel to obtain a convolutional output, which is expressed as: C = σ (W c ·X1 + b c ) Where: C is the convolutional output of the supplier data, σ () is the ReLU activation function, W c is the weight coefficient of the convolutional layer, X1 is the preprocessed supplier data, and b c is the bias coefficient of the convolutional layer.

[0020] The pooling layer is used to perform downsampling on the convolutional output of the supplier data through a max pooling operation to obtain the pooling feature of the supplier data, which is expressed as: P = MaxPool(C) Where: P is the pooling feature of the supplier data, and MaxPool is the max pooling operation.

[0021] The fully connected layer is used to flatten the pooling feature of the supplier data and obtain the mapping feature of the supplier data through the weight coefficient, which is expressed as: F = σ(W f · P + b f ) Where: F is the mapping feature of the supplier data, W f is the weight coefficient of the fully connected layer, b f is the bias coefficient of the fully connected layer.

[0022] The first output layer is used to generate a supplier feature vector based on the mapping feature of the supplier data, expressed as: V s = W₀ · F + b o Where: V s is the supplier feature vector, W₀ is the weight coefficient of the first output layer, b o is the bias coefficient of the first output layer.

[0023] The second input layer is used to receive the second input data, and the second input data is the preprocessed demander data.

[0024] The encoding layer is used to perform non-linear compression on the second input data to obtain the encoded data of the demander data, expressed as: E = σ (W e · X₂ + b e ) Where: E is the encoded data of the demander data, W e is the weight coefficient of the encoding layer, X₂ is the preprocessed demander data, b e is the bias coefficient of the encoding layer.

[0025] The latent space representation layer is used to perform dimensionality reduction and feature extraction on the encoded data of the demander data to generate the low-dimensional feature of the demander data, expressed as: Z = W z · E + b z Where: Z is the low-dimensional feature of the demander data, W z is the weight coefficient of the latent space representation layer, b z is the bias coefficient of the latent space representation layer.

[0026] The decoding layer is used to reconstruct the low-dimensional feature of the demander data to obtain the reconstructed data of the demander data, expressed as: D = σ (W d · Z + b d ) Where: D is the reconstructed data of the demander data, W d is the weight coefficient of the decoding layer, b d is the bias coefficient of the decoding layer.

[0027] The second output layer is used to generate an initial feature vector of the demand side according to the reconstructed data of the demand side data, expressed as: V d =W ′ d ·D + b ′ d Where: V d is the initial feature vector of the demand side, W ′ d is the weight coefficient of the second output layer, b ′ d is the bias coefficient of the second output layer.

[0028] The second output layer normalizes the initial feature vector of the demand side through the Sigmoid function to obtain the feature vector of the demand side.

[0029] The entire process of the scenario-based data model extracts local patterns through a convolutional neural network, and realizes feature dimensionality reduction and scenario-based reconstruction through an autoencoder, and finally forms a semantic alignment feature vector for both the supply and demand sides.

[0030] Specifically, the first output layer obtains the final set of supply-side feature vectors, and the second output layer obtains the set of demand-side feature vectors. The feature vectors in the final set of supply-side feature vectors and the set of demand-side feature vectors include data content features, data context features, and data value features. The data content feature reflects the content attributes of the data itself, including the lexical features of text data and the texture features of image data. The data context feature contains the association information that captures the data in a specific scenario or context, specifically the logical relationship and time sequence between the data. The data value feature is the potential value of the data in the application scenario, including the novelty, scarcity, and practicality of the data.

[0031] S3. Construct a vector knowledge base to store the supply-side feature vectors and demand-side feature vectors after scenario-based reconstruction, and based on the actual demand task and the vector knowledge base, use the supply-demand docking algorithm to obtain the final supply-demand docking result.

[0032] In an optional embodiment of the present invention, the present invention uses the supply-demand docking algorithm to obtain the final supply-demand docking result based on the actual demand task and the vector knowledge base, including the following steps: A1. Based on the vector knowledge base, use the supply-demand docking algorithm to calculate the similarity between the demand side and the supply side, expressed as:

[0033] Where: is the similarity between the demand side and the supply side, is the The feature vector of the supplier after scenario-based reconstruction, is the feature vector of the demander after scenario-based reconstruction, is the magnitude of the feature vector of the supplier after scenario-based reconstruction, is the magnitude of the feature vector of the demander after scenario-based reconstruction.

[0034] A2. Set the similarity threshold between the demander and the supplier based on the actual demand task.

[0035] A3. Obtain the final supply-demand matching result based on the similarity between the demander and the supplier and the similarity threshold between the demander and the supplier.

[0036] Specifically, the present invention determines whether the similarity between the demander and the supplier exceeds the similarity threshold between the demander and the supplier; if so, the corresponding demander and supplier are matched, otherwise the corresponding demander and supplier are not matched, and then the final supply-demand matching result is output.

[0037] In the process of the present invention obtaining the final supply-demand matching result based on the actual demand task and the vector knowledge base by using the supply-demand matching algorithm, an encryption algorithm is used to encrypt the actual demand task and the final supply-demand matching result to ensure the security of data transmission. The encryption algorithm adopted in the present invention is specifically a privacy protection mechanism, which performs access control and permission management on the data to ensure that only authorized users or systems can access and process the data.

[0038] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, and the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0039] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device, and the instruction device implements the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1The functions specified in one or more boxes.

[0040] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide for implementing the steps of the functions specified in one or more processes and / or boxes Figure 1 One process or more processes and / or boxes Figure 1 The steps of the functions specified in one or more boxes.

[0041] In the present invention, specific embodiments are used to elaborate on the principles and implementation manners of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present invention.

[0042] Those of ordinary skill in the art will realize that the embodiments described herein are for helping the reader understand the principles of the present invention, and it should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those of ordinary skill in the art can make various other specific deformations and combinations that do not depart from the essence of the present invention according to the technical revelations disclosed in the present invention, and these deformations and combinations are still within the protection scope of the present invention.

Claims

1. A method for scenario-based supply-demand matching of data elements, characterized in that It includes the following steps: S1. Collect the original data of the supplier and the demander, denoise and format the original data of the supplier and the demander to obtain the preprocessed supplier dataset and demander dataset; S2. Build a scenario-based data model based on the convolutional neural network and the deep autoencoder, and use the scenario-based data model, the preprocessed supplier dataset and demander dataset to obtain the scenario-reconstructed supplier feature vector and demander feature vector; S3. Build a vector knowledge base to store the scenario-reconstructed supplier feature vector and demander feature vector, and based on the actual demand task and the vector knowledge base, use the supply-demand matching algorithm to obtain the final supply-demand matching result.

2. The data element scenario-based supply and demand matching method according to claim 1, wherein In step S2, the scenario-based data model includes a first input layer, a convolutional layer, a pooling layer, a fully connected layer, a first output layer, a second input layer, an encoding layer, a latent space representation layer, a decoding layer, and a second output layer connected in sequence.

3. The data element scenario-based supply-demand matching method according to claim 2, wherein, The data processing process of the scenario-based data model is specifically as follows: The first input layer is used to receive the first input data, and the first input data is the preprocessed supplier data; The convolutional layer is used to extract features from the first input data through a convolutional kernel to obtain a convolutional output, expressed as: C= σ (W c ·X1 + b c ) Where: C is the convolution output of the supplier data, σ ( ) is the ReLU activation function, W c is the weight coefficient of the convolutional layer, X1 is the preprocessed supplier data, b c is the bias coefficient of the convolutional layer; The pooling layer is used to downsample the convolutional output of the supplier data through a max pooling operation to obtain the pooling feature of the supplier data, expressed as: P = MaxPool(C) where: P is the pooling feature of the supplier data, and MaxPool is the max pooling operation; The fully connected layer is used to flatten the pooling feature of the supplier data and obtain the mapping feature of the supplier data through weight coefficients, expressed as: F = σ (W f ·P + b f ) Among them: F is the mapping feature of the supplier data, W f is the weight coefficient of the fully connected layer, b f is the bias coefficient of the fully connected layer; The first output layer is used to generate a supplier feature vector according to the mapping feature of the supplier data, expressed as: V s = W · F + b o Where: V s is the supplier feature vector, and W is the weight coefficient of the first output layer, b o is the bias coefficient of the first output layer; The second input layer is used to receive the second input data, and the second input data is the preprocessed demander data; The encoding layer is used to perform non-linear compression on the second input data to obtain the encoded data of the demander data, expressed as: E= σ (W e ·X2 + b e ) Among them: E is the encoded data of the data of the demanding party, W e is the weight coefficient of the encoding layer, X2 is the data of the demanding party after preprocessing, b e is the bias coefficient of the encoding layer; The latent space representation layer is used to reduce the dimension and extract features from the encoded data of the demander data to generate the low-dimensional feature of the demander data; Z = W z ·E + b z Where: Z is the low-dimensional feature of the demander's data, W z is the weight coefficient of the latent space representation layer, b z is the bias coefficient of the latent space representation layer; The decoding layer is used to reconstruct the low-dimensional feature of the demander data to obtain the reconstructed data of the demander data, expressed as: D= σ (W d ·Z + b d ) Where: D is the reconstructed data of the data of the demander, W d is the weight coefficient of the decoding layer, b d is the bias coefficient of the decoding layer; The second output layer is used to generate an initial demander feature vector according to the reconstructed data of the demander data, expressed as: V d =W ′ d ·D + b ′ d Where: V d is the initial demand-side feature vector, W ′ d is the weight coefficient of the second output layer, b ′ d is the bias coefficient of the second output layer; And normalize the initial demander feature vector through the Sigmoid function to obtain the demander feature vector.

4. The data element scenario-based supply and demand matching method according to claim 1, wherein In step S3, based on the actual demand task and the vector knowledge base, use the supply-demand matching algorithm to obtain the final supply-demand matching result, including the following steps: A1. Based on the vector knowledge base, use the supply-demand matching algorithm to calculate the similarity between the demander and the supplier; A2. Set the similarity threshold between the demander and the supplier based on the actual demand task; A3. Based on the similarity between the demander and the supplier and the similarity threshold between the demander and the supplier, obtain the final supply-demand matching result.

5. The data element scenario-based supply-demand matching method according to claim 4, wherein In step A1, based on the vector knowledge base, use the supply-demand matching algorithm to calculate the similarity between the demander and the supplier, expressed as: Wherein: is the similarity between the demand side and the supply side, is the th supply-side feature vector after scenario reconstruction, is the th demand-side feature vector after scenario reconstruction, is the norm of the th supply-side feature vector after scenario reconstruction, is the norm of the th demand-side feature vector after scenario reconstruction.

6. The data element scenario-based supply and demand matching method according to claim 1, wherein In step S3, in the process of obtaining the final supply-demand matching result by using the supply-demand matching algorithm based on the actual demand task and the vector knowledge base, an encryption algorithm is used to encrypt the actual demand task and the final supply-demand matching result.

Images