Order transaction management method and system based on big data

Abstract

The invention discloses an order transaction management method and system based on big data, and belongs to the technical field of order transaction management. The method comprises the following steps: firstly, obtaining a text description of a user transaction order, then carrying out semantic coding on the text description of the user transaction order to obtain a transaction order semantic feature vector, and then carrying out semantic coding on archive information containing each terminal to obtain a plurality of terminal archive semantic feature vectors; carrying out adaptive memory feature attention fusion on the transaction order semantic feature vector and the plurality of terminal file semantic feature vectors to obtain a plurality of transaction order-terminal semantic association feature vectors, and finally obtaining a plurality of transaction order-terminal semantic association feature vectors based on the plurality of transaction order-terminal semantic association feature vectors. And determining the name and the archive information of the corresponding terminal, so that the terminal suitable for the user can be intelligently recommended according to the user transaction order.

Description

Technical Field

[0001] The present application relates to the field of order transaction management, and more specifically, to an order transaction management method and system based on big data. Background Art

[0002] In traditional building materials transactions, the information flow between manufacturers, distributors and end users is not smooth, which makes it difficult to accurately grasp the market supply and demand relationship, and it is easy to have overcapacity or insufficient supply. At the same time, due to information opacity, distributors find it difficult to obtain real market price information during the procurement and sales process, which increases transaction costs and the possibility of artificial manipulation of building materials prices; the traditional building materials transaction model relies on cumbersome processes such as offline negotiations, contract signing and payment, which is not only time-consuming and labor-intensive, but also increases the operating costs of enterprises. With the development of big data related technologies, the use of big data technology can help building materials companies predict market demand, optimize production plans and inventory management by analyzing historical sales data, market trends, etc., thereby reducing costs and improving efficiency.

[0003] For example, a Chinese patent with application number 201910126515.9 discloses a market material allocation business model system based on construction big data processing, including: a procurement client, a merchant client and a server, wherein the procurement client and the merchant client are connected to the server through a network; the procurement client is used for buyers to publish procurement needs, receive push information from the server and place purchase orders; the merchant client is used for sellers to quote and record material delivery information; the server includes: a data acquisition module, a data processing module, a price calculation module and a push module; by extracting and processing building materials transaction data, the optimal cost of material procurement in multiple shipping locations is calculated, so as to achieve the purpose of reasonably allocating materials in areas with sufficient materials according to cost, avoiding artificial manipulation of building materials prices, and protecting the interests of the purchaser.

[0004] Technologies similar to the above can reasonably adjust the optimal procurement cost based on building big data and avoid artificial manipulation of building material prices. However, in building material transactions, at least the following technical problems still exist. For example, in conventional e-commerce platforms, products are open to all users and have the same price. The only difference for different regions is the logistics and distribution costs. In bulk building materials, such as cement, due to different regions of manufacturers and different transportation routes of raw materials, the same model of goods from different manufacturers will have large price differences. This may lead to cross-regional transactions. It is impossible to intelligently recommend a terminal suitable for the user based on the user's transaction order, which will cause price disorder and is not conducive to maintaining the stability of the building materials trading market.

[0005] Therefore, an optimized order transaction management solution is needed to solve the above technical problems. Summary of the invention

[0006] In order to solve the above technical problems, this application is proposed. The embodiment of this application provides an order transaction management method and system based on big data. It can intelligently recommend a terminal suitable for the user according to the user's transaction order.

[0007] According to one aspect of the present application, a method for managing order transactions based on big data is provided, which includes:

[0008] Get the text description of the user's transaction order;

[0009] Performing semantic encoding on the text description of the user transaction order to obtain a transaction order semantic feature vector;

[0010] Semantically encoding the archive information of each terminal to obtain multiple terminal archive semantic feature vectors;

[0011] The transaction order semantic feature vector and the plurality of terminal archive semantic feature vectors are subjected to an adaptive memory feature attention fusion module to obtain a plurality of transaction order-terminal semantic association feature vectors;

[0012] Based on the multiple transaction order-terminal semantic association feature vectors, the name and profile information of the corresponding terminal are determined.

[0013] Preferably, semantic encoding is performed on the text description of the user's transaction order to obtain a transaction order semantic feature vector, including: data cleaning is performed on the text description of the user's transaction order to obtain a cleaned transaction order description; and the cleaned transaction order description is passed through a transaction order semantic encoder including a word embedding layer to obtain the transaction order semantic feature vector.

[0014] Preferably, the cleaned transaction order description is passed through a transaction order semantic encoder including a word embedding layer to obtain the transaction order semantic feature vector, including: passing the cleaned transaction order description through the word embedding layer to obtain multiple transaction order description word embedding vectors; passing the multiple transaction order description word embedding vectors through the transaction order semantic encoder to obtain the transaction order semantic feature vector.

[0015] Preferably, the multiple transaction order description word embedding vectors are passed through the transaction order semantic encoder to obtain the transaction order semantic feature vector, including: using the transaction order semantic encoder to perform global contextual semantic encoding on the multiple transaction order description word embedding vectors to obtain multiple contextual transaction order feature vectors; cascading the multiple contextual transaction order feature vectors to obtain the transaction order semantic feature vector.

[0016] Preferably, semantic encoding is performed on the archive information containing each terminal to obtain a plurality of terminal archive semantic feature vectors, including: passing the archive information containing each terminal through a terminal archive semantic encoder including a word embedding layer to obtain the plurality of terminal archive semantic feature vectors.

[0017] Preferably, the transaction order semantic feature vector and the multiple terminal archive semantic feature vectors are subjected to an adaptive memory feature attention fusion module to obtain multiple transaction order-terminal semantic association feature vectors, including: inputting the transaction order semantic feature vector and the terminal archive semantic feature vector into a collaborative representation learning module to obtain an implicit collaborative coding vector between transaction order-terminal semantic features; writing the implicit collaborative coding vector between transaction order-terminal semantic features into an adaptive memory module to obtain a transaction order-terminal semantic dynamic key vector; extracting the transaction order-terminal semantic dynamic key vector from the adaptive memory module, and inputting the transaction order semantic feature vector and the terminal archive semantic feature vector into a dynamic feature emphasis module respectively to obtain an optimized transaction order semantic feature vector and an optimized terminal archive semantic feature vector; inputting the optimized transaction order semantic feature vector and the optimized terminal archive semantic feature vector into a feature linear fusion module to obtain the transaction order-terminal semantic association feature vector.

[0018] Preferably, the transaction order semantic feature vector and the terminal archive semantic feature vector are input into a collaborative representation learning module to obtain an implicit collaborative coding vector between transaction order-terminal semantic features, including: adding the transaction order semantic feature vector and the terminal archive semantic feature vector by position to obtain a transaction order-terminal semantic joint representation vector; subtracting the transaction order semantic feature vector and the terminal archive semantic feature vector by position to obtain a transaction order-terminal semantic deviation representation vector; performing positional dot multiplication on the transaction order semantic feature vector and the terminal archive semantic feature vector to obtain a transaction order-terminal semantic element-level interaction representation vector; cascading the transaction order-terminal semantic joint representation vector, the transaction order-terminal semantic deviation representation vector and the transaction order-terminal semantic element-level interaction representation vector to obtain a transaction order-terminal multi-level semantic fusion vector, and performing point convolution on the transaction order-terminal multi-level semantic fusion vector and then inputting it into a leakage corrected linear unit to obtain an implicit collaborative coding vector between transaction order-terminal semantic features.

[0019] Preferably, the transaction order-terminal semantic dynamic key vector is extracted from the adaptive memory module, and the transaction order semantic feature vector and the terminal archive semantic feature vector are respectively input into the dynamic feature emphasis module to obtain the optimized transaction order semantic feature vector and the optimized terminal archive semantic feature vector, including: determining the transaction order semantic query embedding matrix, the transaction order semantic query bias vector, the transaction order semantic value embedding matrix and the transaction order semantic value bias vector; based on the transaction order semantic query embedding matrix and the transaction order semantic query bias vector, performing a first linear transformation on the transaction order semantic feature vector to obtain the transaction order semantic query vector; based on the transaction order semantic value embedding matrix and the transaction order semantic value bias vector, performing a second linear transformation on the transaction order semantic feature vector to obtain the transaction order semantic value vector; and inputting the transaction order-terminal semantic dynamic key vector, the transaction order semantic query vector and the transaction order semantic value vector into the dynamic feature emphasis module based on the converter structure to obtain the optimized transaction order semantic feature vector.

[0020] Preferably, based on the multiple transaction order-terminal semantic association feature vectors, determining the name and profile information of the corresponding terminal includes: passing the multiple transaction order-terminal semantic association feature vectors through a classifier to obtain multiple probability values; returning the name and profile information of the terminal corresponding to the maximum value among the multiple probability values.

[0021] According to another aspect of the present application, there is provided an order transaction management system based on big data, which includes:

[0022] A text description acquisition module is used to obtain the text description of the user's transaction order;

[0023] A text description semantic encoding module, used for semantically encoding the text description of the user's transaction order to obtain a transaction order semantic feature vector;

[0024] The archive information semantic coding module is used to semantically encode the archive information containing each terminal to obtain multiple terminal archive semantic feature vectors;

[0025] A transaction order-terminal semantic association module, used for performing an adaptive memory feature attention fusion module on the transaction order semantic feature vector and the plurality of terminal archive semantic feature vectors to obtain a plurality of transaction order-terminal semantic association feature vectors;

[0026] The matching module is used to determine the name and file information of the corresponding terminal based on the multiple transaction order-terminal semantic association feature vectors.

[0027] This application has at least the following technical effects:

[0028] Compared with the prior art, the big data-based order transaction management method and system provided by the present application first obtains a text description of a user's transaction order, then semantically encodes the text description of the user's transaction order to obtain a transaction order semantic feature vector, then semantically encodes the archive information containing each terminal to obtain multiple terminal archive semantic feature vectors, then performs an adaptive memory feature attention fusion module on the transaction order semantic feature vector and the multiple terminal archive semantic feature vectors to obtain multiple transaction order-terminal semantic association feature vectors, and finally determines the name and archive information of the corresponding terminal based on the multiple transaction order-terminal semantic association feature vectors, so as to intelligently recommend a terminal suitable for the user according to the user's transaction order. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. The following drawings are not intentionally scaled to the actual size, and the focus is on illustrating the main purpose of the present application.

[0030] Figure 1 It is a flowchart of an order transaction management method based on big data according to an embodiment of the present application.

[0031] Figure 2 Schematic diagram of the architecture of the order transaction management method based on big data according to an embodiment of the present application.

[0032] Figure 3 This is a flowchart of sub-step S120 of the order transaction management method based on big data according to an embodiment of the present application.

[0033] Figure 4 This is a flowchart of sub-step S122 of the order transaction management method based on big data according to an embodiment of the present application.

[0034] Figure 5 This is a flowchart of sub-step S150 of the order transaction management method based on big data according to an embodiment of the present application.

[0035] Figure 6 It is a block diagram of an order transaction management system based on big data according to an embodiment of the present application. DETAILED DESCRIPTION

[0036] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work also fall within the scope of protection of the present application.

[0037] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0038] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0039] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0040] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0041] In order to maintain the stability of the building materials trading market, the technical concept of this application is to intelligently recommend a terminal suitable for the user based on the user's trading order. The terminal refers to different manufacturers on the building materials trading platform, specifically the sellers who provide goods.

[0042] Based on this, Figure 1 It is a flowchart of an order transaction management method based on big data according to an embodiment of the present application. Figure 2 FIG. 1 is a schematic diagram of the architecture of the order transaction management method based on big data according to an embodiment of the present application. Figure 1 and Figure 2As shown, according to the big data-based order transaction management method of the embodiment of the present application, the steps include: S110, obtaining a text description of a user's transaction order; S120, semantically encoding the text description of the user's transaction order to obtain a transaction order semantic feature vector; S130, semantically encoding the archive information containing each terminal to obtain a plurality of terminal archive semantic feature vectors; S140, performing an adaptive memory feature attention fusion module on the transaction order semantic feature vector and the plurality of terminal archive semantic feature vectors to obtain a plurality of transaction order-terminal semantic association feature vectors; S150, determining the name and archive information of the corresponding terminal based on the plurality of transaction order-terminal semantic association feature vectors.

[0043] Specifically, in the technical solution of the present application, a text description of a user's transaction order is first obtained. Then, the text description of the user's transaction order is semantically encoded to obtain a transaction order semantic feature vector. Specifically, semantic feature information in the text description of the user's transaction order is extracted, such as the user's geographic location, the type of goods required, and other information, to reflect the user's purchase needs.

[0044] In a specific example of the present application, the encoding process of semantically encoding the text description of the user's transaction order to obtain a transaction order semantic feature vector includes: firstly performing data cleaning on the text description of the user's transaction order to obtain a cleaned transaction order description; and then passing the cleaned transaction order description through a transaction order semantic encoder including a word embedding layer to obtain a transaction order semantic feature vector. In particular, irrelevant information in the transaction order description can be removed by data cleaning.

[0045] More specifically, in an embodiment of the present application, the encoding process of passing the cleaned transaction order description through a transaction order semantic encoder including a word embedding layer to obtain a transaction order semantic feature vector includes: first passing the cleaned transaction order description through the word embedding layer to obtain a plurality of transaction order description word embedding vectors; then, passing the plurality of transaction order description word embedding vectors through the transaction order semantic encoder to obtain the transaction order semantic feature vector.

[0046] At the same time, semantic coding is performed on the archive information containing each terminal to obtain multiple terminal archive semantic feature vectors. Specifically, archive semantic information in the archive information of each terminal is extracted, including the name of the manufacturer, geographical location, product type, price range, etc.

[0047] In a specific example of the present application, semantic encoding of profile information containing each terminal to obtain multiple terminal profile semantic feature vectors is implemented as follows: the profile information containing each terminal is passed through a terminal profile semantic encoder including a word embedding layer to obtain multiple terminal profile semantic feature vectors.

[0048] Next, the transaction order semantic feature vector and the multiple terminal archive semantic feature vectors are subjected to an adaptive memory feature attention fusion module to obtain multiple transaction order-terminal semantic association feature vectors. Specifically, by subjecting the semantic features of the user transaction order to the semantic features of each terminal archive, an association relationship between the two can be constructed to measure the degree of match between the two.

[0049] Accordingly, if Figure 3 As shown, semantic encoding is performed on the text description of the user's transaction order to obtain a transaction order semantic feature vector, including: S121, data cleaning is performed on the text description of the user's transaction order to obtain a cleaned transaction order description; S122, the cleaned transaction order description is passed through a transaction order semantic encoder including a word embedding layer to obtain the transaction order semantic feature vector. It should be understood that the purpose of step S121 is to process and clean the original transaction order text, remove unnecessary information, punctuation marks, special characters, etc., and perform text normalization for subsequent processing and analysis. The cleaned transaction order description is easier to understand and process. The purpose of step S122 is to convert the cleaned transaction order description into a vector representation that can be understood and processed by a computer. The word embedding layer is a model that maps words to a continuous vector space, which can represent each word as a dense vector and capture the semantic relationship between words. The transaction order semantic encoder is a model or algorithm that takes the entire transaction order description as input and converts it into a fixed-length semantic feature vector. This feature vector can be used for subsequent machine learning tasks. In summary, the purpose of steps S121 and S122 is to convert the original transaction order text into a vector representation that can be understood and processed by the machine for subsequent analysis, modeling and application. Data cleaning can improve the quality and consistency of the text and reduce noise and interference, while semantic encoding can capture the semantic information of the text and convert it into numerical features, making subsequent tasks more accurate and effective.

[0050] More specifically, in step S122, Figure 4As shown, the cleaned transaction order description is passed through a transaction order semantic encoder including a word embedding layer to obtain the transaction order semantic feature vector, including: S1221, passing the cleaned transaction order description through the word embedding layer to obtain multiple transaction order description word embedding vectors; S1222, passing the multiple transaction order description word embedding vectors through the transaction order semantic encoder to obtain the transaction order semantic feature vector. It should be understood that word embedding is a technology that maps words to a continuous vector space, which can represent each word as a dense vector and capture the semantic relationship between words. In step S1221, each word in the transaction order description is converted into a corresponding word embedding vector to form a sequence or matrix, which represents the word-level semantic features of the entire transaction order description. Step S1222 passes multiple transaction order description word embedding vectors through a transaction order semantic encoder to obtain a semantic feature vector of the transaction order. The transaction order semantic encoder is a model or algorithm that receives multiple word embedding vectors as input and converts them into a fixed-length semantic feature vector. This feature vector integrates the semantic information of the entire transaction order description and can be used for subsequent machine learning tasks such as classification, clustering, recommendation, etc. The purpose of steps S1221 and S1222 is to convert the cleaned transaction order description into a semantic feature vector of fixed length. Through word embedding, each word is mapped to a vector representation, capturing the semantic relationship between words. Then, the transaction order semantic encoder integrates these word embedding vectors to generate a feature vector representing the semantics of the entire transaction order description. This feature vector can be used to calculate the similarity between texts, perform classification tasks, perform cluster analysis, etc. By converting the transaction order description into a semantic feature vector, machine learning and deep learning methods can be better used for the analysis and application of transaction orders.

[0051] More specifically, in step S1222, the multiple transaction order description word embedding vectors are passed through the transaction order semantic encoder to obtain the transaction order semantic feature vector, including: using the transaction order semantic encoder to perform global contextual semantic encoding on the multiple transaction order description word embedding vectors to obtain multiple contextual transaction order feature vectors; and cascading the multiple contextual transaction order feature vectors to obtain the transaction order semantic feature vector.

[0052] More specifically, in step S130, semantic encoding is performed on the archive information containing each terminal to obtain multiple terminal archive semantic feature vectors, including: passing the archive information containing each terminal through a terminal archive semantic encoder containing a word embedding layer to obtain the multiple terminal archive semantic feature vectors. It should be understood that semantic encoding of the archive information containing each terminal to obtain the semantic feature vectors of multiple terminal archives can bring multiple uses and benefits: 1. Terminal archive analysis: By obtaining the semantic feature vectors of the terminal archive, the terminal archive can be deeply analyzed and understood. These feature vectors can be used to calculate the similarity between terminals, find similar terminal archives, perform cluster analysis, find groups or patterns of terminals, and help enterprises understand and identify different types of terminals. 2. Terminal recommendation: Based on the semantic feature vectors of the terminal archive, personalized terminal recommendations can be performed. By comparing the terminal archive feature vector of the user with the feature vector of the existing terminal archive, the terminal that best matches the user's needs and preferences can be found and recommended. In other words, by semantically encoding the archive information containing each terminal, the semantic feature vectors of multiple terminal archives can be obtained, thereby realizing in-depth analysis, recommendation, classification and identification of the terminal.

[0053] In particular, in the technical solution of the present application, the transaction order semantic feature vector expresses the text semantic features of the text description of the user transaction order, and the terminal archive semantic feature vector expresses the text semantic features of the text description of the terminal archive information. Therefore, after the transaction order semantic feature vector and the terminal archive semantic feature vector are associated and encoded, the text semantic association feature representation of the text semantic features of the transaction order semantic feature vector and the terminal archive semantic feature vector in the full text semantic association space can be expressed. However, while performing the association representation in the full text semantic association space, considering the correspondence difference between the text description of the user transaction order and the text description of the terminal archive information in the source text semantics, the feature distribution discreteness of the transaction order-terminal semantic association feature matrix is ​​high, so that when it is classified by the classifier, the weight matrix of the classifier is difficult to converge with the class label belonging to the predetermined text semantic association space, affecting the training effect of the classifier. Based on this, in a preferred embodiment of the present application, S140, the transaction order semantic feature vector and the multiple terminal archive semantic feature vectors are subjected to an adaptive memory feature attention fusion module to obtain multiple transaction order-terminal semantic association feature vectors. Here, the adaptive memory feature attention fusion module is a new feature interaction fusion architecture, which aims to achieve efficient fusion and enhancement between feature vectors from different sources through multi-stage feature processing and optimization. Specifically, the adaptive memory feature attention fusion module integrates the collaborative representation learning module, the adaptive memory module, the dynamic feature emphasis module based on the converter structure, and the feature linear fusion module to generate a highly optimized and expressive transaction order-terminal semantic association feature vector.

[0054] Specifically, the transaction order semantic feature vector and the terminal archive semantic feature vector are input into the adaptive memory feature attention fusion module to obtain the transaction order-terminal semantic association feature vector, including: inputting the transaction order semantic feature vector and the terminal archive semantic feature vector into the collaborative representation learning module to obtain the implicit collaborative coding vector between the transaction order-terminal semantic features; writing the implicit collaborative coding vector between the transaction order-terminal semantic features into the adaptive memory module to obtain the transaction order-terminal semantic dynamic key vector; extracting the transaction order-terminal semantic dynamic key vector from the adaptive memory module, and inputting the transaction order semantic feature vector and the terminal archive semantic feature vector into the dynamic feature emphasis module respectively to obtain the optimized transaction order semantic feature vector and the optimized terminal archive semantic feature vector; inputting the optimized transaction order semantic feature vector and the optimized terminal archive semantic feature vector into the feature linear fusion module to obtain the transaction order-terminal semantic association feature vector.

[0055] Specifically, the transaction order semantic feature vector and the terminal archive semantic feature vector are input into the collaborative representation learning module to obtain the implicit collaborative coding vector between the transaction order and terminal semantic features, including: adding the transaction order semantic feature vector and the terminal archive semantic feature vector by position to obtain the transaction order-terminal semantic joint representation vector; subtracting the transaction order semantic feature vector and the terminal archive semantic feature vector by position to obtain the transaction order-terminal semantic deviation representation vector; performing point multiplication by position on the transaction order semantic feature vector and the terminal archive semantic feature vector to obtain the transaction order-terminal semantic element-level interaction representation vector; cascading the transaction order-terminal semantic joint representation vector, the transaction order-terminal semantic deviation representation vector and the transaction order-terminal semantic element-level interaction representation vector to obtain the transaction order-terminal multi-level semantic fusion vector, and performing point convolution on the transaction order-terminal multi-level semantic fusion vector and then inputting it into the leakage corrected linear unit to obtain the implicit collaborative coding vector between the transaction order-terminal semantic features. Specifically, the process can be expressed by the formula:

[0056]

[0057] Wherein, V1 represents the semantic feature vector of the transaction order, V2 represents the semantic feature vector of the terminal file, Indicates adding by position point. represents positional subtraction, ⊙ represents positional multiplication, concat represents the cascade function, Conv 1×1 represents point convolution, Leaky ReLU represents leaky rectified linear unit, V J Represents the implicit collaborative encoding vector between transaction order and terminal semantic features.

[0058] Specifically, the transaction order semantic feature vector and the terminal archive semantic feature vector are first input into a collaborative representation learning module, which consists of a point convolution layer and an activation function, and is designed to extract high-level abstract features from the original feature vector. Through a carefully designed network architecture, the collaborative representation learning module is able to capture the deep association between the two feature vectors, the so-called "implicit synergy". This implicit synergy is not limited to simple linear combinations, but involves complex nonlinear interaction patterns. For example, through convolution operations, the network can identify local association patterns, and the use of activation functions, such as ReLU or Leaky ReLU, ensures that the network can learn nonlinear feature representations, which is crucial for capturing complex relationships between features.

[0059] Specifically, the implicit collaborative coding vector between the transaction order and the terminal semantic features is written into the adaptive memory module to obtain the transaction order and the terminal semantic dynamic key vector

[0060] m u =M(V J )

[0061] Among them, V J represents the implicit collaborative coding vector between transaction order and terminal semantic features, M(·) represents a 1×1 convolution operation, and m u Represents the transaction order-terminal semantic dynamic key vector.

[0062] Specifically, the implicit collaborative coding vector between the features is written into an adaptive memory module to obtain a dynamic key vector. The adaptive memory module is a special type of memory structure, which works similarly to the memory unit in a recursive neural network, but has greater flexibility and adaptability. The adaptive memory module can dynamically adjust its own memory content according to the current input state. This means that when a new feature vector is input, the adaptive memory module can adjust the memory according to the newly input data to better adapt to the context requirements of the current scene.

[0063] Specifically, the transaction order-terminal semantic dynamic key vector is extracted from the adaptive memory module, and the transaction order semantic feature vector and the terminal archive semantic feature vector are respectively input into the dynamic feature emphasis module to obtain the optimized transaction order semantic feature vector and the optimized terminal archive semantic feature vector, including: determining the transaction order semantic query embedding matrix, the transaction order semantic query bias vector, the transaction order semantic value embedding matrix and the transaction order semantic value bias vector; based on the transaction order semantic query embedding matrix and the transaction order semantic query bias vector, performing a first linear transformation on the transaction order semantic feature vector to obtain the transaction order semantic query vector; based on the transaction order semantic value embedding matrix and the transaction order semantic value bias vector, performing a second linear transformation on the transaction order semantic feature vector to obtain the transaction order semantic value vector; and inputting the transaction order-terminal semantic dynamic key vector, the transaction order semantic query vector and the transaction order semantic value vector into the dynamic feature emphasis module based on the converter structure to obtain the optimized transaction order semantic feature vector.

[0064]

[0065] Among them, m u Represents the transaction order-terminal semantic dynamic key vector, W 1q is the transaction order semantic query embedding matrix, b 1q is the transaction order semantic query bias vector, V 1q is the transaction order semantic query vector, W 1v is the transaction order semantic value embedding matrix, b 1v is the transaction order semantic value bias vector, V1v is the transaction order semantic value vector, m u T is the transposed vector of the transaction order-terminal semantic dynamic key vector, S is the length of the transaction order-terminal semantic dynamic key vector, softmax represents the normalized exponential function, represents matrix multiplication, V′1 represents the optimized transaction order semantic feature vector, W 2q is the terminal archive semantic query embedding matrix, b 2q is the terminal profile semantic query bias vector, V 2q is the terminal archive semantic query vector, W 2v is the terminal profile semantic value embedding matrix, b 2v is the terminal file semantic value bias vector, V 2v is the terminal file semantic value vector, and V′2 represents the optimized terminal file semantic feature vector.

[0066] Specifically, the dynamic key vector is extracted from the adaptive memory module, and the transaction order semantic feature vector and the dynamic key vector are input into the dynamic feature emphasis module to obtain the optimized transaction order semantic feature vector. At the same time, the dynamic key vector is extracted from the adaptive memory module, and the terminal archive semantic feature vector and the dynamic key vector are input into the dynamic feature emphasis module to obtain the optimized terminal archive semantic feature vector. The core of the dynamic feature emphasis module is the self-attention mechanism, which allows the model to consider the influence of all other elements when processing each input element, not just the adjacent elements. This global perspective helps the model capture long-distance dependencies, which is particularly important for understanding complex data. In the technical solution of the present application, the dynamic feature emphasis module can be used to perform query attention-based feature modulation on the transaction order semantic feature vector and the terminal archive semantic feature vector based on the dynamic key vector, which is equivalent to using the dynamic key vector as a common attention space to not only enhance the expression of key information in the feature vector, but also reduce the influence of noise and irrelevant information, while also making the optimized transaction order semantic feature vector and the optimized terminal archive semantic feature vector aligned in the semantic space, and utilizing the subsequent feature interaction and fusion between the two.

[0067] Specifically, the optimized transaction order semantic feature vector and the optimized terminal archive semantic feature vector are input into a feature linear fusion module to obtain the transaction order-terminal semantic association feature vector.

[0068] V c =αV′1+(1-α)V′2

[0069] Among them, V′1 represents the optimized transaction order semantic feature vector, V′2 represents the optimized terminal archive semantic feature vector, α represents the weight hyperparameter, V c Represents the transaction order-terminal semantic association feature vector.

[0070] Specifically, the optimized transaction order semantic feature vector and the optimized terminal profile semantic feature vector are input into the feature linear fusion module to obtain the transaction order-terminal semantic association feature vector. The main task of the feature linear fusion module is to further integrate the two optimized feature vectors through linear transformation to generate the final transaction order-terminal semantic association feature vector.

[0071] In summary, the adaptive memory feature attention fusion module is elucidated, which can effectively capture the implicit synergistic relationship between feature vectors, and use the implicit synergistic relationship between feature vectors as the key vector to construct the same attention space to perform attention modulation optimization on the transaction order semantic feature vector and the terminal archive semantic feature vector in real time, thereby emphasizing important information and suppressing noise and redundant information, and aligning the transaction order semantic feature vector and the terminal archive semantic feature vector, and finally linearly fuse the optimized transaction order semantic feature vector and the optimized terminal archive semantic feature vector to not only maintain the unique information of each feature vector, but also generate a comprehensive transaction order-terminal semantic association feature vector, providing high-quality input for subsequent tasks.

[0072] Furthermore, the multiple transaction order-terminal semantic association feature vectors are passed through a classifier to obtain multiple probability values; and the name and file information of the terminal corresponding to the maximum value of the multiple probability values ​​are returned. Specifically, the classifier can give a probability value between the transaction order and the terminal file according to the magnitude of the semantic similarity to characterize the possibility of a match between the two. By returning the terminal name and file information corresponding to the maximum probability value, the user can be provided with the best building materials transaction terminal, which can not only ensure that the user can buy suitable goods, but also protect the interests of manufacturers and maintain market stability.

[0073] Accordingly, if Figure 5 As shown, based on the multiple transaction order-terminal semantic association feature vectors, determining the name and profile information of the corresponding terminal, includes: S151, passing the multiple transaction order-terminal semantic association feature vectors through a classifier to obtain multiple probability values; S152, returning the name and profile information of the terminal corresponding to the maximum value among the multiple probability values.

[0074] It should be understood that the role of the classifier is to use the given categories and known training data to learn classification rules and classifiers, and then classify (or predict) unknown data. Logistic regression and SVM are often used to solve binary classification problems. For multi-class classification problems, logistic regression or SVM can also be used, but multiple binary classifications are required to form a multi-classification. However, this is prone to errors and is not efficient. Commonly used multi-classification methods include the Softmax classification function.

[0075] In summary, the big data-based order transaction management method according to the embodiment of the present application is explained, which can intelligently recommend a terminal suitable for the user based on the user's transaction order.

[0076] Figure 6 FIG. 1 is a block diagram of an order transaction management system 100 based on big data according to an embodiment of the present application. Figure 6 As shown, according to the embodiment of the present application, the order transaction management system 100 based on big data includes: a text description acquisition module 110, which is used to obtain the text description of the user's transaction order; a text description semantic encoding module 120, which is used to semantically encode the text description of the user's transaction order to obtain a transaction order semantic feature vector; an archive information semantic encoding module 130, which is used to semantically encode the archive information containing each terminal to obtain a plurality of terminal archive semantic feature vectors; a transaction order-terminal semantic association module 140, which is used to perform an adaptive memory feature attention fusion module on the transaction order semantic feature vector and the plurality of terminal archive semantic feature vectors to obtain a plurality of transaction order-terminal semantic association feature vectors; a matching module 150, which is used to determine the name and archive information of the corresponding terminal based on the plurality of transaction order-terminal semantic association feature vectors.

[0077] Here, those skilled in the art can understand that the specific functions and operations of each module in the above-mentioned order transaction management system based on big data 100 have been referred to above. Figures 1 to 5 The description of the order transaction management method based on big data has been introduced in detail, and therefore, its repeated description will be omitted.

[0078] As described above, the order transaction management system 100 based on big data according to the embodiment of the present application can be implemented in various wireless terminals, such as a server with an order transaction management algorithm based on big data. In one example, the order transaction management system 100 based on big data according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the order transaction management system 100 based on big data can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the order transaction management system 100 based on big data can also be one of the many hardware modules of the wireless terminal.

[0079] Alternatively, in another example, the big data-based order transaction management system 100 and the wireless terminal may also be separate devices, and the big data-based order transaction management system 100 may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0080] In addition, it will be appreciated by those skilled in the art that various aspects of the present application may be illustrated and described by a number of patentable categories or situations, including any new and useful process, machine, product or combination of substances, or any new and useful improvements thereto. Accordingly, various aspects of the present application may be performed entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. The above hardware or software may all be referred to as "data blocks", "modules", "engines", "units", "components" or "systems". In addition, various aspects of the present application may be represented as a computer product located in one or more computer-readable media, which includes computer-readable program code.

[0081] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the invention belongs. It should also be understood that terms such as those defined in common dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology and should not be interpreted in an idealized or extremely formal sense, unless explicitly defined as such herein.

[0082] The above is an explanation of the present invention and should not be considered as a limitation thereof. Although several exemplary embodiments of the present invention have been described, it will be readily appreciated by those skilled in the art that many modifications may be made to the exemplary embodiments without departing from the novel teachings and advantages of the present invention. Therefore, all such modifications are intended to be included within the scope of the present invention as defined in the claims. It should be understood that the above is an explanation of the present invention and should not be considered as being limited to the specific embodiments disclosed, and modifications to the disclosed embodiments and other embodiments are intended to be included within the scope of the appended claims. The present invention is defined by the claims and their equivalents.

Claims

1. An order transaction management method based on big data, characterized in that: include: Get the text description of the user's transaction order; Performing semantic encoding on the text description of the user transaction order to obtain a transaction order semantic feature vector; Semantically encoding the archive information of each terminal to obtain multiple terminal archive semantic feature vectors; The transaction order semantic feature vector and the plurality of terminal archive semantic feature vectors are subjected to an adaptive memory feature attention fusion module to obtain a plurality of transaction order-terminal semantic association feature vectors; Based on the multiple transaction order-terminal semantic association feature vectors, the name and profile information of the corresponding terminal are determined.

2. The order transaction management method based on big data according to claim 1 is characterized in that: Semantically encoding the text description of the user transaction order to obtain a transaction order semantic feature vector includes: Performing data cleaning on the text description of the user's transaction order to obtain a cleaned transaction order description; The cleaned transaction order description is passed through a transaction order semantic encoder including a word embedding layer to obtain the transaction order semantic feature vector.

3. The order transaction management method based on big data according to claim 2 is characterized in that: The cleaned transaction order description is passed through a transaction order semantic encoder including a word embedding layer to obtain the transaction order semantic feature vector, including: Passing the cleaned transaction order description through the word embedding layer to obtain a plurality of transaction order description word embedding vectors; The multiple transaction order description word embedding vectors are passed through the transaction order semantic encoder to obtain the transaction order semantic feature vector.

4. The order transaction management method based on big data according to claim 3 is characterized in that: Passing the multiple transaction order description word embedding vectors through the transaction order semantic encoder to obtain the transaction order semantic feature vector includes: Using the transaction order semantic encoder, performing global context semantic encoding on the multiple transaction order description word embedding vectors to obtain multiple context transaction order feature vectors; The multiple contextual transaction order feature vectors are cascaded to obtain the transaction order semantic feature vector.

5. The order transaction management method based on big data according to claim 4 is characterized in that: Semantic encoding is performed on the archive information of each terminal to obtain multiple terminal archive semantic feature vectors, including: The archive information including each terminal is passed through a terminal archive semantic encoder including a word embedding layer to obtain the plurality of terminal archive semantic feature vectors.

6. The order transaction management method based on big data according to claim 5 is characterized in that: The transaction order semantic feature vector and the plurality of terminal archive semantic feature vectors are subjected to an adaptive memory feature attention fusion module to obtain a plurality of transaction order-terminal semantic association feature vectors, including: Inputting the transaction order semantic feature vector and the terminal profile semantic feature vector into a collaborative representation learning module to obtain an implicit collaborative encoding vector between transaction order and terminal semantic features; Writing the implicit collaborative coding vector between the transaction order-terminal semantic features into the adaptive memory module to obtain a transaction order-terminal semantic dynamic key vector; Extract the transaction order-terminal semantic dynamic key vector from the adaptive memory module, and input the transaction order semantic feature vector and the terminal profile semantic feature vector into the dynamic feature emphasis module to obtain an optimized transaction order semantic feature vector and an optimized terminal profile semantic feature vector; The optimized transaction order semantic feature vector and the optimized terminal archive semantic feature vector are input into a feature linear fusion module to obtain the transaction order-terminal semantic association feature vector.

7. The order transaction management method based on big data according to claim 6 is characterized in that: Inputting the transaction order semantic feature vector and the terminal archive semantic feature vector into a collaborative representation learning module to obtain an implicit collaborative encoding vector between transaction order and terminal semantic features, including: Adding the transaction order semantic feature vector and the terminal archive semantic feature vector by position to obtain a transaction order-terminal semantic joint representation vector; Subtracting the transaction order semantic feature vector from the terminal file semantic feature vector by position to obtain a transaction order-terminal semantic deviation representation vector; Performing position point multiplication on the transaction order semantic feature vector and the terminal profile semantic feature vector to obtain a transaction order-terminal semantic element-level interaction representation vector; The transaction order-terminal semantic joint representation vector, the transaction order-terminal semantic deviation representation vector and the transaction order-terminal semantic element-level interaction representation vector are cascaded to obtain a transaction order-terminal multi-level semantic fusion vector, and the transaction order-terminal multi-level semantic fusion vector is point-convolved and then input into a leakage-corrected linear unit to obtain an implicit collaborative coding vector between transaction order-terminal semantic features.

8. The order transaction management method based on big data according to claim 7 is characterized in that: Extracting the transaction order-terminal semantic dynamic key vector from the adaptive memory module, and inputting the transaction order semantic feature vector and the terminal archive semantic feature vector into the dynamic feature emphasis module respectively to obtain an optimized transaction order semantic feature vector and an optimized terminal archive semantic feature vector, including: Determine a transaction order semantic query embedding matrix, a transaction order semantic query bias vector, a transaction order semantic value embedding matrix, and a transaction order semantic value bias vector; Based on the transaction order semantic query embedding matrix and the transaction order semantic query bias vector, performing a first linear transformation on the transaction order semantic feature vector to obtain a transaction order semantic query vector; Based on the transaction order semantic value embedding matrix and the transaction order semantic value bias vector, performing a second linear transformation on the transaction order semantic feature vector to obtain a transaction order semantic value vector; The transaction order-terminal semantic dynamic key vector, the transaction order semantic query vector and the transaction order semantic value vector are input into a dynamic feature emphasis module based on a converter structure to obtain an optimized transaction order semantic feature vector.

9. The order transaction management method based on big data according to claim 8 is characterized in that: Determining the name and profile information of the corresponding terminal based on the multiple transaction order-terminal semantic association feature vectors includes: Passing the plurality of transaction order-terminal semantic association feature vectors through a classifier to obtain a plurality of probability values; The name and profile information of the terminal corresponding to the maximum value among the multiple probability values ​​are returned.

10. The order transaction management system based on big data is characterized by: include: A text description acquisition module is used to obtain the text description of the user's transaction order; A text description semantic encoding module, used for semantically encoding the text description of the user's transaction order to obtain a transaction order semantic feature vector; The archive information semantic coding module is used to semantically encode the archive information containing each terminal to obtain multiple terminal archive semantic feature vectors; A transaction order-terminal semantic association module, used for performing an adaptive memory feature attention fusion module on the transaction order semantic feature vector and the plurality of terminal archive semantic feature vectors to obtain a plurality of transaction order-terminal semantic association feature vectors; The matching module is used to determine the name and file information of the corresponding terminal based on the multiple transaction order-terminal semantic association feature vectors.

Images