Component recommendation method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention discloses a component recommendation method and device, electronic equipment and a storage medium, and the method comprises the steps: responding to a component recommendation operation triggered by a target user; obtaining a component dependence network graph, component demand description information of the target user and historical interaction information of each historical interaction component used by the target user in the current editing item; determining a plurality of to-be-screened components from a component library on the basis of the component dependency network graph and the historical interaction information; based on the component dependency network diagram, the component attribute information of each component in the component library, the component demand description information and each piece of historical interaction information, determining the adaptation attribute of each component relative to the target user; according to the method, the to-be-screened components are screened on the basis of the adaptive attributes, the target recommended component is obtained, the component search efficiency and modeling fluency are remarkably improved, the burden of repeatedly trying different component search conditions by a user is reduced, and finally the efficiency and quality of the overall modeling process are optimized.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a component recommendation method, apparatus, electronic device, and storage medium. Background Technology

[0002] With the rapid development of computer technology, simulation technology plays an increasingly important role in engineering research and development. Modelica-based modeling and simulation platforms, with their powerful physical modeling capabilities and cross-disciplinary characteristics, have become the mainstream tool for simulating complex multidisciplinary systems. These platforms, by providing rich component libraries, support users in quickly building high-precision simulation models, significantly improving the efficiency and quality of engineering design.

[0003] Currently, users mainly use the platform's built-in search function to find the components they need in the component library, relying on keyword matching or simple classification and filtering mechanisms.

[0004] However, as the component library continues to expand, traditional search methods suffer from the following problems: Firstly, when users cannot accurately describe the target component, it is difficult to provide precise search results. Secondly, users often need to repeatedly try different search criteria to find components that match their needs. Therefore, existing component search methods suffer from low efficiency, severely impacting the smoothness of the overall modeling process. This not only prolongs the user's search time but also reduces modeling efficiency. Summary of the Invention

[0005] This invention provides a component recommendation method, apparatus, electronic device, and storage medium to improve component search efficiency and modeling fluency, reduce the burden on users to repeatedly try different component search conditions, and ultimately optimize the efficiency and quality of the overall modeling process.

[0006] In a first aspect, embodiments of the present invention provide a component recommendation method, the method comprising:

[0007] In response to a component recommendation operation triggered by a target user, the system obtains a component dependency network graph, the target user's component requirement description information, and the historical interaction information of each historical interactive component used by the target user in the current editing project; wherein, each node in the component dependency network graph corresponds to a different component in the component library, and the connecting edges between nodes represent the association relationship between components;

[0008] Based on the component dependency network graph and the historical interaction information, multiple components to be screened are determined from the component library;

[0009] Based on the component dependency network graph, the component attribute information of each component in the component library, the component requirement description information, and the historical interaction information, the adaptation attributes of each component relative to the target user are determined.

[0010] Based on the aforementioned adaptation attributes, the plurality of components to be filtered are processed to obtain at least one target recommendation component corresponding to the target user.

[0011] Secondly, embodiments of the present invention also provide a component recommendation device, the device comprising:

[0012] The component recommendation response module is used to respond to the component recommendation operation triggered by the target user, and to obtain the component dependency network graph, the component requirement description information of the target user, and the historical interaction information of each historical interactive component used by the target user in the current editing project; wherein, each node in the component dependency network graph corresponds to a different component in the component library, and the connecting edges between nodes represent the association relationship between components;

[0013] The component to be filtered module is used to determine multiple components to be filtered from the component library based on the component dependency network graph and the historical interaction information.

[0014] The adaptation attribute determination module is used to determine the adaptation attribute of each component relative to the target user based on the component dependency network graph, the component attribute information of each component in the component library, the component requirement description information, and the historical interaction information of each historical interaction component.

[0015] The target component recommendation module is used to filter the plurality of components to be filtered based on each of the adaptation attributes, so as to obtain at least one target recommended component corresponding to the target user.

[0016] Thirdly, embodiments of the present invention also provide an electronic device, the electronic device comprising:

[0017] One or more processors;

[0018] A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the component-recommended method as described in any embodiment of the present invention.

[0019] Fourthly, embodiments of the present invention also provide a storage medium containing computer-executable instructions, which, when executed by a computer processor, are used to perform a component recommendation method as described in any of the embodiments of the present invention.

[0020] The technical solution of this invention, in response to a component recommendation operation triggered by a target user, acquires a component dependency network graph, the target user's component requirement description information, and historical interaction information of each historical interactive component used by the target user in the current editing project. Based on the component dependency network graph and the historical interaction information, multiple components to be filtered are determined from the component library. Then, based on the component dependency network graph, the component attribute information of each component in the component library, the component requirement description information, and the historical interaction information of each historical interactive component, the adaptation attributes of each component relative to the target user are determined. Finally, based on the adaptation attributes, the multiple components to be filtered are processed to obtain at least one target recommended component corresponding to the target user. The technical solution provided in this embodiment mines the potential relationships between components through a component dependency network graph, encodes the user's functional requirements by combining component attribute information and requirement description information, and models user behavior preferences using historical interaction information. This enables a more accurate understanding of user needs. Even when the user cannot accurately describe the target component, it can dynamically recommend suitable components based on the dependencies between components and the user's historical behavior, thereby significantly improving component search efficiency and modeling smoothness, reducing the burden on users to repeatedly try different search conditions, and ultimately optimizing the efficiency and quality of the overall modeling process. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of exemplary embodiments of the present invention, the accompanying drawings used in describing the embodiments are briefly introduced below. Obviously, the accompanying drawings described are only a portion of the drawings of the embodiments to be described in this invention, and not all of the drawings. For those skilled in the art, other drawings can be obtained from these drawings without any creative effort.

[0022] Figure 1 This is a flowchart illustrating a component recommendation method provided in an embodiment of the present invention;

[0023] Figure 2 This is a flowchart illustrating yet another component recommendation method provided in an embodiment of the present invention;

[0024] Figure 3 This is a schematic diagram of the node embedding and propagation process involved in an embodiment of the present invention;

[0025] Figure 4 This is a schematic diagram illustrating the execution logic of the component recommendation method involved in the embodiments of the present invention;

[0026] Figure 5 This is a schematic diagram of the structure of a component recommendation device provided in an embodiment of the present invention;

[0027] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0028] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, the accompanying drawings show only the parts relevant to the present invention, and not all of the structures.

[0029] Example 1

[0030] Figure 1 This is a flowchart illustrating a component recommendation method provided in an embodiment of the present invention. This embodiment can be applied to any situation in which suitable components are intelligently recommended to the user during the simulation modeling process. The method can be executed by a component recommendation device, which can be implemented in the form of software and / or hardware. The hardware can be an electronic device, such as a mobile terminal, a PC, or a server.

[0031] like Figure 1 As shown, the recommended methods for this component include:

[0032] S110. In response to the component recommendation operation triggered by the target user, obtain the component dependency network diagram, the target user's component requirement description information, and the historical interaction information of each historical interactive component used by the target user in the current editing project.

[0033] The target user refers to a specific user who actively triggers the component recommendation function in the current modeling or simulation project. Their identity is characterized by using a modeling platform based on the Modelica language for system design. The component recommendation operation is a behavioral instruction initiated by the target user during the modeling process, requesting component recommendations. This operation serves as the trigger condition for starting the recommendation process, indicating that the user needs assistance in selecting components suitable for the current modeling scenario. Optionally, the component recommendation operation can be implemented through specific interface interactions, such as clicking a recommendation button; or implicit behaviors, such as inputting a requirement description.

[0034] In this graph, each node corresponds to a different component in the component library, and the edges connecting nodes represent the relationships between components. The component dependency network graph is a mathematical model that represents the relationships between components in a component library in a graph structure. Nodes represent specific components in the library, while edges characterize functional dependencies, data transfer, or logical connections between components. This graph structure systematically describes the collaborative working patterns between components through topological connections. It can include both explicit interface compatibility relationships and implicit coupling relationships at the engineering application level, providing a quantitative basis for the strength of relationships between components in recommendation algorithms. It is the fundamental data structure for calculating component suitability and recommendation priority.

[0035] It's important to note that a component can be a single component or a component class. In other words, the elements corresponding to nodes in the component dependency network graph can be either specific component entities with independent functions within the component library, or logical collections (component classes) composed of multiple components with similar functions or parameters. This dual representation allows the network graph to describe both fine-grained relationships between specific components and macroscopic relationships between component categories. This enables recommendation systems to perform component relationship analysis and recommendation result generation at different levels of abstraction, satisfying both users' precise needs for specific components and adapting to fuzzy search scenarios involving broad component categories.

[0036] Component requirement descriptions refer to the specific requirements of target users for the functions and characteristics of the required components, expressed to the system through natural language, keywords, or structured input. These descriptions can include objective attributes such as component technical parameters and performance indicators, as well as subjective aspects such as application scenarios and design intent.

[0037] The current editing project refers to a specific engineering task or simulation model that the target user is using the modeling platform for design and development. It includes the model structure the user has built within this project environment, the set of components selected, and the relevant engineering parameter configurations. Historical interaction components refer to the set of components that the target user has actually used in the current editing project. These components record the specific interaction behaviors generated by the user with the system during the design process through operations such as selection, configuration, or connection. Historical interaction information refers to the collection of a series of operational behavior data generated by the target user on each historical interaction component in the current editing project, including but not limited to the component selection frequency, usage duration, parameter configuration records, and operation traces such as connection relationships with other components.

[0038] In this embodiment, the component recommendation method can be encapsulated as an independent software functional module, which can be called a target component recommendation tool, and integrated into the existing modeling platform system, making it one of the platform's built-in functions. When a user performs a specific component recommendation operation in the modeling platform, the tool will automatically trigger and execute the following process: First, it retrieves a pre-built component dependency network diagram that reflects the relationships between components. Then, it obtains the component requirement description information input by the user, as well as the historical interaction information of each historical interaction component in the current project. The historical interaction information contains detailed operation data. This information together constitutes the input data basis of the target component recommendation tool, providing multi-dimensional basis for subsequent component selection and suitability evaluation.

[0039] For example, User A develops a car powertrain model using the Modelica-based modeling platform. When User A is on the modeling interface, they can enter the search keyword "high-efficiency turbocharger" in the interactive interface of the target component recommendation tool and click the "component recommendation" control. This triggers the component recommendation operation, and the integrated target component recommendation tool will immediately perform the following operations: First, it calls the component dependency network graph pre-stored on the platform, which contains the connection relationship between the engine component and the transmission component. Then, it retrieves the search keyword "high-efficiency turbocharger" that the user just entered. At the same time, it collects the components used in the project, such as "2.0L four-cylinder engine" and "6-speed automatic transmission," and their usage records, such as the engine parameters being modified multiple times and the transmission being frequently connected to other components. These data together constitute the input basis for intelligent recommendation.

[0040] S120. Based on the component dependency network graph and historical interaction information, identify multiple components to be screened from the component library.

[0041] Among them, the components to be screened refer to the set of potential candidate components initially selected from the component library. These components not only meet the basic condition of being related to the components already used by the user, but also meet the preliminary matching requirements of the current project context. They are the basic candidate set for subsequent adaptability assessment and personalized ranking. Their number is usually larger than the final recommendation result but much smaller than the size of the complete component library, reflecting the phased processing strategy of the recommendation tool from coarse screening to fine screening.

[0042] Specifically, by analyzing the topological relationships between components in the component dependency network graph and combining this with the user's historical interaction records of components used in the current project, a set of candidate components related to the user's current modeling needs can be initially screened from the complete component library. Specifically, firstly, adjacent components with functional associations to historically interacted components are identified based on the connection relationships between nodes in the network graph. Then, considering historical behavioral data such as interaction frequency and usage scenarios, a subset of components that are relatively well-matched in terms of structural relevance and usage habits is selected, forming a candidate pool for subsequent refined recommendations.

[0043] S130. Based on the component dependency network diagram, the component attribute information of each component in the component library, the component requirement description information, and the historical interaction information of each historical interaction component, determine the adaptation attributes of each component relative to the target user.

[0044] The comprehensive matching index is calculated for each component to be screened. This index quantitatively reflects the degree to which the component matches the current modeling needs of the target user. Optionally, the adaptability attribute can be characterized by a component adaptability score. The higher the component adaptability score of a component to be screened, the higher the degree to which the component matches the current modeling needs of the target user; the lower the component adaptability score of a component to be screened, the lower the degree to which the component matches the current modeling needs of the target user.

[0045] In this embodiment, firstly, the component dependency network graph is used to analyze the topological adaptability and connection strength of components in the overall architecture; secondly, the matching degree between the functional characteristics and user needs is evaluated by combining component attribute information; then, the component requirement description information is parsed to capture the user's explicitly expressed needs and intentions; finally, historical interaction information is integrated to mine the user's implicit usage preferences and behavioral patterns, and finally, these heterogeneous features are fused into a comprehensive importance quantification index through a specific weighting algorithm or machine learning model. This index directly reflects the degree of adaptability of each component to the user's current modeling needs.

[0046] More specifically, the determination of the adaptation attributes of each component relative to the target user can be implemented in two stages: The first stage establishes preference attributes that reflect the user's personalized component selection tendency by analyzing the structural characteristics of the component dependency network, the adaptability of component attributes, the semantics of requirement description, and historical interaction behavior. This information quantifies the user's attention weight and selection probability for different component features. The second stage inputs the preference attributes and the topological characteristics of the component dependency network into a predefined component adaptability calculation model. This model comprehensively evaluates the adaptability of components to individual users through a specific calculation process, and finally outputs a comprehensive importance score for each component relative to the target user. This score considers both the functional position of the component in the engineering system and the personalized needs of the user.

[0047] S140. Based on each adaptation attribute, filter multiple components to be filtered to obtain at least one target recommended component corresponding to the target user.

[0048] The target recommendation component refers to the set of components that are ultimately selected through the recommendation method and are highly matched with the current modeling needs of the target user. It is the final output of the recommendation process. The number of target recommendation components can be one or more.

[0049] Specifically, the candidate component set can be optimized and filtered using a preset filtering strategy based on the pre-calculated adaptation attribute values ​​of each component to be filtered. This filtering strategy may include, but is not limited to, threshold filtering, sorting and truncation filtering, and / or clustering analysis filtering. For example, all components to be filtered can first be prioritized or categorized based on their adaptation attribute scores. Then, using preset recommendation rules, such as Top-N selection or dynamic threshold filtering, one or more target recommended components that meet the filtering criteria can be extracted. Finally, a list of recommended results that best matches the target user's current project context and personalized needs can be generated. This list contains one or more highly adaptable components as a reference for user decision-making.

[0050] For example, the target component recommendation tool has calculated the compatibility attribute values ​​(0-1 points) of 10 components to be screened (such as battery modules, motor controllers, etc.). Among them, "lithium-ion battery pack" scores 0.92 points and "IGBT inverter" scores 0.88 points. If 0.85 points is set as the recommendation threshold, the tool will automatically select the two components with the highest scores as the target recommended components and display their 3D models and parameter comparisons on the interface. At the same time, it will mark "best compatibility with your current project" as a smart prompt to help users make quick decisions.

[0051] The technical solution of this embodiment of the invention, in response to a component recommendation operation triggered by a target user, obtains a component dependency network graph, the target user's component requirement description information, and the historical interaction information of each historical interactive component used by the target user in the current editing project. Based on the component dependency network graph and the historical interaction information, multiple components to be filtered are determined from the component library. Furthermore, based on the component dependency network graph, the component attribute information of each component in the component library, the component requirement description information, and the historical interaction information of each historical interactive component, the adaptation attributes of each component relative to the target user are determined. Finally, based on the adaptation attributes, the multiple components to be filtered are processed to obtain at least one target recommended component corresponding to the target user. The technical solution provided in this embodiment mines the potential relationships between components through a component dependency network graph, encodes the user's functional requirements by combining component attribute information and requirement description information, and models user behavior preferences using historical interaction information. This enables a more accurate understanding of user needs. Even when the user cannot accurately describe the target component, it can dynamically recommend suitable components based on the dependencies between components and the user's historical behavior, thereby significantly improving component search efficiency and modeling smoothness, reducing the burden on users to repeatedly try different search conditions, and ultimately optimizing the efficiency and quality of the overall modeling process.

[0052] Example 2

[0053] Figure 2This diagram illustrates a component recommendation method provided in an embodiment of the present invention. Based on the foregoing embodiments, steps S120 and 130 are further refined. Specific implementation details can be found in the technical solution of this embodiment. Technical terms that are the same as or corresponding to those in the above embodiments will not be repeated here.

[0054] like Figure 2 As shown, the method specifically includes the following steps:

[0055] S210. In response to the component recommendation operation triggered by the target user, obtain the component dependency network diagram, the target user's component requirement description information, and the historical interaction information of each historical interactive component used by the target user in the current editing project.

[0056] In this component dependency network graph, each node corresponds to a different component in the component library, and the connecting edges between nodes represent the association between components.

[0057] S220. Based on the component identification information of historical interaction information, determine the neighbor nodes corresponding to each historical interaction component from the component dependency network graph.

[0058] Component identification information refers to key metadata used to uniquely identify and locate a specific component in the component dependency network graph. It contains an identifier system that can accurately point to a specific component or component class in the component library. Neighbor nodes are adjacent nodes in the component dependency network graph that have direct edges connected to the nodes corresponding to historically interacted components. These nodes represent other components that have explicit functional associations, data interactions, or logical dependencies with the components already used by the user. The type and weight of their connection edges reflect the degree of collaboration between different components.

[0059] In this embodiment, by parsing the unique component identifiers (i.e., component identification information) recorded in historical interaction information, the node positions corresponding to each historical component used by the user are accurately located in the component dependency network graph. Then, based on the topology of the network graph, all directly connected adjacent nodes of these nodes are extracted. The advantage of this setup is that it maps the user's actual operation records to structured path queries in the network graph, and discovers a set of potential candidate components that have direct engineering connections with the user's existing components through graph traversal techniques, providing an initial filtering range based on network relationships for subsequent recommendations.

[0060] S230. Determine the components corresponding to each neighbor node as multiple components to be filtered.

[0061] In this embodiment, the physical components corresponding to each neighbor node can be aggregated into a pool of components to be screened. This pool retains the topological association features in the original network graph and has specific engineering objects that can be used for attribute analysis and functional evaluation, providing a candidate range that is both structured and semantic for subsequent adaptation calculation.

[0062] S240. Based on the component dependency network diagram, the component attribute information of each component in the component library, the component requirement description information, and the historical interaction information of each historical interaction component, determine the target user's preference attributes for different components.

[0063] Among them, preference attributes refer to personalized indicators that are quantified and reflect the target user's preference for specific components or component features.

[0064] As an alternative implementation method, these heterogeneous data can be input into a specific preference analysis model to quantify the user's preference weights for the functional characteristics, brand type, or parameter range of various components, and finally output a quantifiable preference attribute matrix. This matrix reflects both the technical characteristics of the user's explicit needs and the implicit selection preferences extracted from behavioral data.

[0065] In another implementation, S240 may include the following steps:

[0066] S1. Based on the component requirement description information and the historical interaction information of each historical interaction component, determine the user behavior characteristics.

[0067] Among them, user behavior characteristics refer to the quantitative feature vectors that reflect the user's component usage habits and preferences by analyzing the user's UI component requirement description information and the target user's component interaction records in the current project, such as operation data including selection frequency, usage duration, and parameter configuration.

[0068] In this embodiment, user behavior features reflecting personalized information are constructed by fusing and parsing the explicit component requirement descriptions and implicit historical interaction behavior data of the target user. Specifically, this can include: first, using natural language processing techniques to extract structured features such as keywords and functional requirements from the requirement descriptions; and second, using behavioral analysis algorithms to mine operational patterns such as component usage frequency, parameter adjustment patterns, and combinations of related components from historical interaction information. Then, these two types of features are fused into a unified low-dimensional feature vector using feature engineering methods. This vector represents both the user's explicit requirements for the current project and their inherent behavioral patterns formed over long-term use, ultimately forming a quantifiable and calculable representation of user behavior features.

[0069] Optionally, the specific implementation steps of S1 may include:

[0070] (1) The user requirement description information is encoded based on the preset encoder to obtain the user requirement description encoding vector.

[0071] The pre-trained encoder refers to a pre-trained feature extraction model whose function is to transform unstructured raw input data (such as text descriptions) into a machine-processable numerical vector representation. The user requirement description encoding vector refers to the conversion of component requirement description information input by the user in natural language form into a fixed-dimensional numerical vector representation through the pre-trained encoder.

[0072] Specifically, a pre-trained language feature extraction model can be used to transform unstructured user-inputted requirement description text into a machine-processable numerical feature representation. First, the original natural language text undergoes word segmentation and vectorization preprocessing. Then, it is input into a neural network encoder with a fixed architecture (such as a Transformer, LSTM, or BERT encoder). This encoder extracts the syntactic structure and semantic features of the text through multiple layers of nonlinear transformations, ultimately outputting a low-dimensional, dense real-number vector. This vector retains the key semantic information of the original requirement description in the feature space while eliminating the diversity and redundancy of language expression, forming a standardized numerical representation that includes both explicit requirement features and implicit potential intentions, providing the basic input for subsequent deep feature fusion.

[0073] (2) Based on a two-layer fully connected network, feature extraction is performed on the user demand description encoding vector to obtain user demand description features.

[0074] Among them, the user demand description feature refers to the high-order semantic representation obtained by deep feature refinement of the original user demand description encoding vector through a two-layer fully connected neural network.

[0075] Specifically, the user demand description encoding vector can be input into a neural network consisting of two consecutive fully connected layers for deep feature transformation. The first fully connected layer performs dimensionality upscaling mapping of the feature space through a non-linear activation function, capturing the high-order feature combinations implicit in the encoding vector; the second fully connected layer performs dimensionality reduction, removing redundant information and retaining the most discriminative semantic features. After the cascaded processing of these two layers, the original encoding vector is transformed into a low-dimensional dense vector with stronger representational power. This vector retains the core semantics of user demands and optimizes the feature distribution through adaptive learning of the neural network, forming a demand feature representation more suitable for subsequent recommendation tasks.

[0076] For example, this step can be represented as: encoding the user requirement description into a vector E. mThe input is fed into a fully connected network (FFN). This layer of the network performs linear transformations and non-linear activation operations on the input vector, mapping the original embedding vector to a new feature space, thereby obtaining the user's u-embedded representation (i.e., user demand description features):

[0077]

[0078] in, This represents the final user demand description features. Here, a two-layer fully connected network is used, so l is set to 2, and σ is the ReLU activation function.

[0079] (3) For the historical interaction information of each historical interaction component, the encoding vector of the historical interaction component is obtained based on the preset encoder.

[0080] Historical interaction information includes interaction frequency, usage duration, and component parameter configuration records for historical interaction components. In essence, historical interaction information refers to the core behavioral data generated during user interaction with components, encompassing at least three aspects: First, interaction frequency quantifies the number of times a user selects the component, reflecting its basic usage frequency; second, usage duration records the length of time a user continuously operates on the component, reflecting its criticality; and third, component parameter configuration records preserve the user's history of adjusting component technical parameters, revealing personalized configuration preferences. This structured data collectively constitutes the underlying data source for user behavior analysis. Through multi-dimensional features such as frequency statistics, duration distribution, and parameter change patterns, it comprehensively characterizes the depth of user interaction and customization tendencies with components, providing fine-grained raw behavioral features for subsequent encoding vector generation. The historical interaction component encoding vector refers to the numerical representation obtained after feature extraction from the user-component interaction records using a preset encoder.

[0081] In this embodiment, a pre-trained preset encoder model can be used to transform the multi-dimensional interaction records corresponding to each historical interaction component into a unified numerical vector representation. The preset encoder maps discrete interaction frequencies into dense vectors through a feature embedding layer, while simultaneously processing continuous duration and parameter data through a neural network. Finally, all interaction features are fused into a single fixed-length vector. This vector retains the statistical characteristics of the original interaction behavior and extracts higher-order behavior patterns through the encoder's nonlinear transformation, forming a standardized mathematical expression that simultaneously reflects user operating habits and component usage characteristics, providing a basic input for subsequent behavioral feature analysis. Optionally, the preset encoder used in this step can be a neural network encoder with a fixed architecture (such as a Transformer, LSTM, or BERT encoder).

[0082] (4) Based on a two-layer fully connected network, feature extraction is performed on the encoding vector of the historical interaction component to obtain the features of the historical interaction component.

[0083] Among them, the historical interaction component features refer to the high-order behavioral pattern representation obtained by deep feature refinement of the original historical interaction component encoding vector through a two-layer fully connected neural network.

[0084] In this embodiment, the feature extraction process of the historical interaction component features is consistent with the specific implementation of step (2), which can be referred to in step (2) and will not be repeated here.

[0085] (5) The average value of the features of each historical interactive component is calculated to obtain the features of the historical interactive components to be spliced.

[0086] In this embodiment, the feature vectors of multiple historical interaction components are aggregated into a single vector by adding them element-wise and then taking the average. After dimensional alignment of all historical interaction component features, an averaging operation is performed on each dimension of the feature space. This linear aggregation method preserves the common patterns of features of each component while smoothing out abnormal feature values ​​of individual components, ultimately generating a composite feature vector that integrates the user's overall interaction history, which is the feature vector of the historical interaction components to be concatenated. This vector not only retains the frequency distribution information of the user's use of different components, but also strengthens the smoothing of noisy data and weakens the impact of outliers through averaging, forming a feature representation of the user's comprehensive behavioral preferences to be concatenated, providing a global historical behavioral context for subsequent fusion with the feature of the requirement description.

[0087] (6) The user demand description features and the historical interaction component features to be spliced ​​are spliced ​​together to obtain user behavior features.

[0088] In this embodiment, the process of combining the feature vector reflecting the user's current explicit needs and the feature vector representing historical interaction components to be concatenated, through a vector concatenation operation, is carried out along the feature dimension. Specifically, first, the dimensions of the two feature vectors are ensured to match, and then they are directly connected along the feature axis to form a higher-dimensional composite vector. This concatenation method not only preserves the independent information structure of the original features but also establishes the correlation between the semantics of needs and behavioral patterns. The resulting user behavior feature vector simultaneously encodes the user's immediate needs and long-term behavioral preferences, providing a comprehensive feature representation that is both real-time and personalized for subsequent recommendation algorithms.

[0089] S2. Based on the component dependency network graph and the component attribute information of each component in the component library, determine the component attribute characteristics.

[0090] Among them, component attribute features refer to the numerical feature vectors that reflect the technical characteristics and functional correlations of components, extracted by analyzing the topology of the component dependency network graph and the inherent attribute information of each component in the component library.

[0091] In this embodiment, a feature vector that comprehensively represents the characteristics of a component can be constructed by fusing the topological features of the component dependency network graph with the detailed attribute data recorded in the component library.

[0092] Optionally, S2 may specifically include the following steps:

[0093] (1) The component attribute information of each component in the component library is encoded based on the preset encoder to obtain the component attribute encoding vector corresponding to each component.

[0094] Among them, the component attribute encoding vector refers to the low-dimensional dense vector representation obtained by extracting features and numerically transforming the original attribute data (such as technical parameters, functional descriptions and other structured information) of each component in the component library through a preset encoder.

[0095] In this embodiment, the specific implementation method for determining the corresponding component attribute encoding vector is consistent for each component in the component library. For ease of calculation, the component attribute set can be constructed from the component attribute information corresponding to each component. in, This represents the component attribute information of the i-th component. Next, this set of component attributes is input into a preset encoder, for example, a BERT encoder, to obtain the component attribute encoding vector corresponding to each component. Among them, E c It is a second-order tensor, E c ∈R k*l k is the total number of components in the component library, and l is the word vector dimension of the BERT pre-trained model, which is 768.

[0096] (2) Configure the component attribute encoding vector corresponding to each component as the node parameter information of the corresponding component in the component dependency network graph to obtain the component attribute network graph.

[0097] Among them, the component attribute network graph refers to an enhanced graph data structure that combines topological structure and node features, formed by using the component attribute encoding vector corresponding to each node as the feature attribute of the node, based on the original component dependency network graph.

[0098] In this embodiment, this step mainly involves fusing the encoded component attribute feature vectors with the topology of the component dependency network graph. Specifically, the encoded vectors of each component attribute are dynamically bound as feature attributes to the corresponding component nodes in the component dependency network graph. This upgrades the original graph structure, which only contained connection relationships, into an enhanced network graph where nodes possess both structural connection attributes and deep semantic features. This configuration operation achieves a unified representation of component static attributes and system-level associations, preserving the structured connection information of the original network while injecting the functional characteristics of the components through node parameter information. This provides complete input data with both topological and semantic features for subsequent feature extraction from the graph neural network.

[0099] (3) Input the component attribute network graph into the pre-trained graph neural network model for feature extraction to obtain the component attribute features corresponding to each component.

[0100] Among them, the graph neural network model is used to extract high-order feature representations that integrate topological relationships and node attributes from the component attribute network graph.

[0101] In this embodiment, this step mainly involves inputting the network graph, which incorporates component attribute encoding vectors, into a pre-trained graph neural network (GNN) model. Deep feature learning is then performed through multi-layer graph convolution or graph attention operations. Based on a message-passing mechanism, the GNN iteratively aggregates the attribute features of each node with the topological information of its neighboring nodes. While preserving the original component functional attributes, it dynamically learns the contextual association features of the components within the network structure, ultimately outputting a high-order feature representation of each component node enhanced by the graph structure. These features include both the component's static attributes and its dynamic association patterns within the system, forming a composite feature vector that comprehensively reflects the component's technical characteristics and engineering application scenarios, providing a deeply optimized component representation for subsequent preference calculations.

[0102] An example is provided below: The embedding representation of a node is updated by propagating information from neighboring nodes, thereby enabling feature extraction of the component node. See the diagram illustrating the node embedding propagation process. Figure 3 .

[0103] For propagation at the k-th layer, given the embedding of the current node i, The embedding propagation layer will compute a new embedding representation for the node. This allows nodes to gradually incorporate more structural information. In this module, the embedding propagation layer employs the GraphSAGE (Graph Sample and Aggregation) algorithm, which randomly samples neighbors and aggregates their features to capture the relationships between nodes within the graph structure. The embedding propagation process will be described in detail below.

[0104] For any node i, in the embedding propagation at layer k, m neighbors are first randomly sampled to obtain all embedding representations of the neighbor set N(i) of node i. (where j∈N(i)). Next, we use mean aggregation to average the embeddings of these neighboring nodes to obtain the aggregated information of m neighboring nodes:

[0105]

[0106] After aggregating the features of m neighbors, the embedding of node i itself is concatenated with the aggregated neighbor embeddings to form a new vector. Here Indicates the splicing operation:

[0107]

[0108] The concatenated embedding vector will be passed through a weight matrix W (k) Perform a linear transformation to generate the final embedding of node i. To enhance the nonlinear expressive power of the model, an activation function σ is usually added to the result after the linear transformation. Here, the ReLU activation function is used:

[0109]

[0110] here This is the updated embedded representation of node i.

[0111] Finally, the feature vector corresponding to each graph node output by the graph neural network model is determined as the component attribute feature corresponding to each component.

[0112] Optionally, when training the graph neural network, the KL divergence loss function can be used to evaluate the model quality. For better model training, the loss function used is as follows:

[0113]

[0114] In the formula, the loss function mainly consists of two parts: the first part measures the loss calculated by the weights, and the second part is the L2 regularization term, used to avoid overfitting of the model. Here, P(i) is the true distribution, Q(i) is the predicted distribution, and θ... i These are model hyperparameters.

[0115] S3. Perform a dot product operation on the component attribute features and user behavior features to obtain the preference attribute vector.

[0116] Each element in the preference attribute vector represents the target user's preference attribute for different components.

[0117] In this embodiment, the two feature vectors are first aligned dimensionally. Then, a dot product operation is performed on them, multiplying them element-wise and summing the results. This operation essentially calculates the similarity projection of the two vectors in the feature space, and the result quantifies the degree of matching between user behavior patterns and component characteristics. In the generated preference attribute vector, each element value represents the adaptation score between the target user and the corresponding component. The higher the value, the more the component matches the user's functional needs and operating habits. These scores constitute the core basis for subsequent recommendation ranking, realizing a mathematical mapping from the feature space to the degree of personalized preference.

[0118] For example, the dimension of the component attribute features is [node_num, dim], which is fixed; the dimension of the user behavior features is [1, dim]. Performing a dot product operation on the component attribute features and user behavior features yields a preference attribute vector with a dimension of [1, node_num]. Optionally, since the target output of this step is personalized weights, and these personalized weights must satisfy a probability distribution, a softmax operation is performed on the final output. Assuming there are 990 components in the component library, for the target user, the preference attribute vector has a dimension of [1, 990], where each element represents the user's preference for each component node.

[0119] S250. Based on the component dependency network graph, preference attributes, and component adaptability operation model, determine the adaptability attributes of each component relative to the target user.

[0120] The component suitability computation model refers to an improved graph ranking algorithm used in component recommendation tools. This model combines the traditional PageRank algorithm with personalized preference attributes, iteratively calculating the importance score of a component in the network graph.

[0121] Specifically, the topology of the component-dependent network graph is first used as the basic transition matrix, and the preference attribute vector is incorporated as the personalized jump probability. Based on the inherent correlation strength between components, components with user preferences are given higher propagation weights. Finally, the output is an adaptation attribute score that reflects both the system-level importance of the component and the user's personalized preferences, so as to achieve a comprehensive quantitative evaluation of the component recommendation priority from two dimensions: the global network structure and the individual user characteristics.

[0122] Optionally, the specific implementation of S250 may include:

[0123] (1) Based on the component dependency network graph, determine the transition matrix that participates in the iterative operation of the component adaptability operation model.

[0124] The transition matrix is ​​a probability transition matrix used in the component adaptability operation model to describe the relationship between components. Its matrix element values ​​are determined by the topology of the component dependency network graph, and each element quantifies the transition probability of jumping from one component to another.

[0125] In this embodiment, the connection strength of each edge in the component-dependent network graph is first quantified. Then, the outgoing edge weights of each node are normalized to ensure that the sum of the elements in each row is 1, thereby constructing a random matrix that meets the requirements of a Markov chain. Each element of this transition matrix represents the probability of transitioning from one component to another during the algorithm iteration process. This preserves the connection structure characteristics of the original network while satisfying the mathematical norms of probability operations, providing a basic probability transition rule for subsequent personalized PageRank algorithms.

[0126] Specifically, the transition matrix can be represented as:

[0127]

[0128] In the formula, M ij Let A represent the value of the element in the i-th row and j-th column of the transition matrix. ij This represents the connection between node i and node j. For example, in a component dependency network graph, if there is an edge pointing from node i to node j, then Aij = 1; otherwise, Aij = 0. j The out-degree of node j, and N is the total number of nodes.

[0129] (2) Use the target user’s preference attributes for different components as the initialization vector of the component adaptability operation model.

[0130] The initialization vector, in this context, is a column vector representing the importance distribution of each component in the initial state of the algorithm during the iterative calculation of the component adaptability computation model. Its element values ​​are obtained by normalizing the preference attribute vector, reflecting the target user's initial preference for different components. This vector serves as prior knowledge injection into the personalized recommendation algorithm, ensuring that the iteration process evolves from an initial state that aligns with user preferences. Its element sum of 1 satisfies the requirements of the probability distribution, preserving user-specific characteristics while providing standardized initial conditions for subsequent iteration convergence. It is a key input parameter for achieving personalized recommendation results.

[0131] In this embodiment, the preference attribute vector representing the user's personalized preferences can be normalized and used as the initial probability distribution vector for the iterative calculation of the component adaptability computation model. Specifically, the preference attribute vector can be used as the importance probability distribution of each component in the initial state of the algorithm, so that the subsequent iteration process, while considering the network topology, is always guided by the user's initial preferences, ultimately achieving a balanced integration of the global importance of the network and the user's personalized needs.

[0132] (3) Based on the transition matrix and initialization vector, perform iterative calculations on the component adaptability calculation model until the offset between the current calculation result and the previous calculation result of the component adaptability calculation model meets the preset conditions, and then determine the current calculation result as the target calculation result.

[0133] In this embodiment, the PageRank component adaptability computation model is used to repeatedly perform matrix operations on the transition matrix and the initialization vector. Each iteration dynamically integrates the network topology propagation effect with the user's initial preferences. In each iteration, the following calculations are performed: multiplying the current state vector by the transition matrix to achieve probability propagation of the network structure, while retaining a portion of the initialization vector to maintain user preferences, generating a new state vector; continuously monitoring the Euclidean distance or cosine similarity between adjacent iterations, terminating the computation when the offset value is less than a preset convergence threshold. At this point, the state vector has reached a stable equilibrium, and its element values ​​reflect both the global importance of the component in the network and the user's personalized preferences. Finally, the comprehensive adaptation score of each component after network structure adjustment is output as the target computation result.

[0134] Specifically, the component adaptability computation model can be represented as:

[0135] r (t+1) =αMr (t) +(1-α)·Pu

[0136] In the formula, M is the transition matrix, Pu represents the initialization vector, α is the damping coefficient, usually set to 0.85, representing the probability that the user continues to select a component, (1-α) represents the probability of randomly selecting any component, and r (t+1) r represents the result of the current operation. (t) This indicates the result of the previous operation.

[0137] (4) Based on the target calculation results, determine the adaptation attributes of each component relative to the target user.

[0138] In this embodiment, the final output of the component adaptability calculation model (i.e., the target calculation result) is used to quantitatively evaluate the degree of matching between each component in the component library and the target user's needs and behavioral characteristics. Finally, a set of numerical adaptation attributes is generated, where each value represents the comprehensive probability or weight of the corresponding component meeting user needs in dimensions such as functional relevance, suitability for usage scenarios, or continuity of historical interactions, thereby providing an objective basis for subsequent screening.

[0139] Optionally, each element value in the target calculation result can be determined as the adaptation attribute of the corresponding component relative to the target user. This can be achieved by performing numerical transformation on the elements in the target calculation result and using the final result as the adaptation attribute of the corresponding component relative to the target user.

[0140] S260. Based on each adaptation attribute, filter multiple components to be filtered to obtain at least one target recommended component corresponding to the target user.

[0141] The following example illustrates the component recommendation method provided in this embodiment of the invention. Figure 4 This is a schematic diagram illustrating the execution logic of the component recommendation method. The component recommendation method provided in this embodiment of the invention can be integrated into a toolkit and can be loaded into any graphics editing software. Figure 4 As shown, during the method development phase, a component dependency network graph is generated in advance based on a large number of users' component interaction records over a historical period. Furthermore, the personalized weight calculation model is trained and validated based on these historical user component interaction records and component description information, resulting in a trained personalized weight calculation model. In this embodiment, the personalized weight calculation model's role is to: identify multiple components to be screened from the component library based on the component dependency network graph and historical interaction information; and determine the suitability attributes of each component relative to the target user based on the component dependency network graph, component attribute information, component requirement description information, and historical interaction information. In practical application, when the target user triggers a component recommendation operation, the component dependency network graph, the target user's component requirement description information, and the historical interaction information of each historical interaction component used by the target user in the current editing project can be obtained. Then, by inputting the obtained information into the trained personalized weight calculation model, the target user's preference attributes for different components can be obtained. Thus, based on the component dependency network graph, preference attributes, and component suitability calculation model (i.e., ... Figure 4 The PageRank module (in the CSS) determines the adaptation attributes of each component relative to the target user. Finally, multiple components to be filtered can be processed based on each adaptation attribute to obtain at least one target recommended component corresponding to the target user.

[0142] The technical solution of this invention, when determining the adaptation attributes of each component relative to the target user, can determine the target user's preference attributes between different components based on the component dependency network graph, component attribute information of each component in the component library, component requirement description information, and historical interaction information of each historically interacting component. Thus, based on the component dependency network graph, preference attributes, and component adaptability calculation model, the adaptation attributes of each component relative to the target user are determined. The technical solution provided in this embodiment firstly mines the potential relationships between components using the component dependency network graph, ensuring that the recommendation results conform to the rationality of the technical logic and system architecture; secondly, it encodes the user's functional requirements using component attribute information and requirement description information, enhancing the relevance of the recommended content; finally, it models user behavior preferences through historical interaction information, making the recommendation results adapt to the user's usage habits. This not only improves the accuracy of the recommendation but also dynamically balances the user's immediate needs and long-term behavioral patterns through iterative optimization of the component adaptability calculation model, thereby achieving the optimal trade-off between technical feasibility, functional adaptability, and user experience.

[0143] Example 3

[0144] Figure 5 This is a schematic diagram of a component recommendation device provided in an embodiment of the present invention. The device includes: a component recommendation response module 310, a component to be screened determination module 320, an adaptation attribute determination module 330, and a target component recommendation module 340.

[0145] The component recommendation response module 310 is used to respond to the component recommendation operation triggered by the target user, and to obtain the component dependency network diagram, the component requirement description information of the target user, and the historical interaction information of each historical interactive component used by the target user in the current editing project; wherein, each node in the component dependency network diagram corresponds to a different component in the component library, and the connecting edges between nodes represent the association relationship between components;

[0146] The component selection module 320 is used to select multiple components from the component library based on the component dependency network graph and historical interaction information.

[0147] The adaptation attribute determination module 330 is used to determine the adaptation attributes of each component relative to the target user based on the component dependency network diagram, the component attribute information of each component in the component library, the component requirement description information, and the historical interaction information of each historical interaction component.

[0148] The target component recommendation module 340 is used to filter multiple components to be filtered based on various adaptation attributes, so as to obtain at least one target recommended component corresponding to the target user.

[0149] The technical solution of this embodiment of the invention, in response to a component recommendation operation triggered by a target user, obtains a component dependency network graph, the target user's component requirement description information, and the historical interaction information of each historical interactive component used by the target user in the current editing project. Based on the component dependency network graph and the historical interaction information, multiple components to be filtered are determined from the component library. Furthermore, based on the component dependency network graph, the component attribute information of each component in the component library, the component requirement description information, and the historical interaction information of each historical interactive component, the adaptation attributes of each component relative to the target user are determined. Finally, based on the adaptation attributes, the multiple components to be filtered are processed to obtain at least one target recommended component corresponding to the target user. The technical solution provided in this embodiment mines the potential relationships between components through a component dependency network graph, encodes the user's functional requirements by combining component attribute information and requirement description information, and models user behavior preferences using historical interaction information. This enables a more accurate understanding of user needs. Even when the user cannot accurately describe the target component, it can dynamically recommend suitable components based on the dependencies between components and the user's historical behavior, thereby significantly improving component search efficiency and modeling smoothness, reducing the burden on users to repeatedly try different search conditions, and ultimately optimizing the efficiency and quality of the overall modeling process.

[0150] Based on the above device, optionally, the component to be screened determination module 320 is specifically used to determine the neighbor node corresponding to each historical interaction component from the component dependency network graph based on the component identification information of the historical interaction information; and to determine the component corresponding to each neighbor node as multiple components to be screened.

[0151] Based on the above-mentioned device, optionally, the adaptation attribute determination module 330 includes:

[0152] The preference attribute determination submodule is used to determine the target user's preference attributes among different components based on the component dependency network graph, the component attribute information of each component in the component library, the component requirement description information, and the historical interaction information of each historical interaction component.

[0153] The important attribute determination submodule is used to determine the adaptation attributes of each component relative to the target user based on the component dependency network graph, the preference attributes, and the component adaptability operation model.

[0154] Based on the above-mentioned device, an optional preference attribute determination submodule includes:

[0155] The user behavior feature determination unit is used to determine user behavior features based on the component requirement description information and the historical interaction information of each of the historical interaction components.

[0156] The component attribute feature determination unit is used to determine component attribute features based on the component dependency network graph and the component attribute information of each component in the component library;

[0157] The preference attribute determination unit is used to perform a dot product operation on the component attribute features and the user behavior features to obtain a preference attribute vector; wherein each element in the preference attribute vector is used to characterize the target user's preference attribute for different components.

[0158] Based on the above-mentioned device, optionally, a user behavior feature determination unit is specifically used to encode the user demand description information based on a preset encoder to obtain a user demand description encoding vector; to extract features from the user demand description encoding vector based on a two-layer fully connected network to obtain user demand description features; to obtain a historical interaction component encoding vector based on the historical interaction information of each of the historical interaction components using a preset encoder; wherein, the historical interaction information includes interaction frequency information and interaction component attribute information corresponding to the historical interaction component; to extract features from the historical interaction component encoding vector based on a two-layer fully connected network to obtain historical interaction component features; to perform an average operation on each of the historical interaction component features to obtain historical interaction component features to be concatenated; and to concatenate the user demand description features and the historical interaction component features to be concatenated to obtain user behavior features.

[0159] Based on the above-mentioned device, optionally, a component attribute feature determination unit is used to encode the component attribute information of each component in the component library based on a preset encoder to obtain a component attribute encoding vector corresponding to each component; configure the component attribute encoding vector corresponding to each component as the node parameter information of the corresponding component in the component dependency network graph to obtain a component attribute network graph; input the component attribute network graph into a pre-trained graph neural network model for feature extraction to obtain the component attribute features corresponding to each component.

[0160] Based on the above-mentioned device, optionally, an important attribute determination submodule is used to: determine the transition matrix participating in the iterative operation of the component adaptability operation model based on the component dependency network graph; use the target user's preference attributes between different components as the initialization vector of the component adaptability operation model; perform iterative operation on the component adaptability operation model according to the transition matrix and the initialization vector until the offset value between the current operation result and the previous operation result of the component adaptability operation model meets a preset condition, and determine the current operation result as the target operation result; and determine the adaptability attribute of each component relative to the target user based on the target operation result.

[0161] The component recommendation device provided in this embodiment of the invention can execute the component recommendation method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method execution.

[0162] It is worth noting that the various units and modules included in the above system are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy differentiation and are not used to limit the protection scope of the embodiments of the present invention.

[0163] Example 4

[0164] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Figure 6 The electronic device 40 shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of the present invention.

[0165] like Figure 6 As shown, electronic device 40 is represented in the form of a general-purpose computing device. The components of electronic device 40 may include, but are not limited to: one or more processors or processing units 401, system memory 402, and bus 403 connecting different system components (including system memory 402 and processing unit 401).

[0166] Bus 403 represents one or more of several bus architectures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the various bus architectures. Examples of these architectures include, but are not limited to, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MAC) bus, the Enhanced ISA bus, the Video Electronics Standards Association (VESA) local bus, and the Peripheral Component Interconnect (PCI) bus.

[0167] Electronic device 40 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by electronic device 40, including volatile and non-volatile media, removable and non-removable media.

[0168] System memory 402 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 404 and / or cache memory 405. Electronic device 40 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 406 may be used to read and write non-removable, non-volatile magnetic media (… Figure 6 Not shown; usually referred to as a "hard drive"). Although Figure 6Not shown, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk") and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 403 via one or more data media interfaces. Memory 402 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of the present invention.

[0169] A program / utility 408 having a set (at least one) of program modules 407 may be stored, for example, in memory 402. Such program modules 407 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. Program modules 407 typically perform the functions and / or methods described in the embodiments of the present invention.

[0170] Electronic device 40 can also communicate with one or more external devices 409 (e.g., keyboard, pointing device, display 810, etc.), and with one or more devices that enable a user to interact with electronic device 40, and / or with any device that enables electronic device 40 to communicate with one or more other computing devices (e.g., network card, modem, etc.). This communication can be performed via input / output (I / O) interface 411. Furthermore, electronic device 40 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 412. As shown, network adapter 412 communicates with other modules of electronic device 40 via bus 403. It should be understood that, although... Figure 6 Not shown, other hardware and / or software modules may be used in conjunction with electronic device 40, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0171] The processing unit 401 executes various functional applications and page processing by running programs stored in the system memory 402, such as implementing the component recommendation method provided in the embodiments of the present invention.

[0172] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via I / O interface 411, or installed from storage system 406. When the computer program is executed by processing unit 401, it performs the functions defined in the methods of the embodiments of the present invention.

[0173] Example 5

[0174] This invention also provides a storage medium containing computer-executable instructions, which, when executed by a computer processor, are used to perform a component recommendation method, including:

[0175] In response to a component recommendation operation triggered by a target user, the system obtains a component dependency network graph, the target user's component requirement description information, and the historical interaction information of each historical interactive component used by the target user in the current editing project; wherein, each node in the component dependency network graph corresponds to a different component in the component library, and the connecting edges between nodes represent the association relationship between components;

[0176] Based on the component dependency network graph and the historical interaction information, multiple components to be screened are determined from the component library;

[0177] Based on the component dependency network graph, the component attribute information of each component in the component library, the component requirement description information, and the historical interaction information of each historical interaction component, the adaptation attributes of each component relative to the target user are determined.

[0178] Based on the aforementioned adaptation attributes, the plurality of components to be filtered are processed to obtain at least one target recommendation component corresponding to the target user.

[0179] The computer storage medium of this invention can be any combination of one or more computer-readable media. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0180] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0181] Program code contained on a computer-readable medium may be transmitted using any suitable medium, including—but not limited to—wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0182] Computer program code for performing the operations of embodiments of the present invention can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages—such as Java, Smalltalk, and C++—as well as conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0183] Note that the above description is merely a preferred embodiment of the present invention and the technical principles employed. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and may include many other equivalent embodiments without departing from the concept of the present invention, the scope of which is determined by the scope of the appended claims.

Claims

1. A component recommendation method, characterized in that, include: In response to a component recommendation operation triggered by a target user, the system obtains a component dependency network graph, the target user's component requirement description information, and the historical interaction information of each historical interactive component used by the target user in the current editing project; wherein, each node in the component dependency network graph corresponds to a different component in the component library, and the connecting edges between nodes represent the association relationship between components; Based on the component dependency network graph and the historical interaction information, multiple components to be screened are determined from the component library; Based on the component dependency network graph, the component attribute information of each component in the component library, the component requirement description information, and the historical interaction information, the adaptation attributes of each component relative to the target user are determined. Based on the aforementioned adaptation attributes, the plurality of components to be filtered are processed to obtain at least one target recommendation component corresponding to the target user.

2. The method according to claim 1, characterized in that, Based on the component dependency network graph and the historical interaction information, multiple components to be screened are determined from the component library, including: Based on the component identification information in the historical interaction information, the neighbor nodes corresponding to each historical interaction component are determined from the component dependency network graph; The components corresponding to each of the neighboring nodes are identified as multiple components to be filtered.

3. The method according to claim 1, characterized in that, The step of determining the adaptation attributes of each component relative to the target user based on the component dependency network graph, the component attribute information of each component in the component library, the component requirement description information, and the historical interaction information includes: Based on the component dependency network graph, the component attribute information of each component in the component library, the component requirement description information, and the historical interaction information of each historical interaction component, the target user's preference attributes for different components are determined. Based on the component dependency network graph, the preference attributes, and the component adaptability operation model, the adaptability attributes of each component relative to the target user are determined.

4. The method according to claim 3, characterized in that, The step of determining the target user's preference attributes for different components based on the component dependency network graph, the component attribute information of each component in the component library, the component requirement description information, and the historical interaction information of each historical interaction component includes: Based on the component requirement description information and the historical interaction information of each of the historical interaction components, user behavior characteristics are determined. Based on the component dependency network graph and the component attribute information of each component in the component library, the component attribute characteristics are determined; Perform a dot product operation on the component attribute features and the user behavior features to obtain a preference attribute vector; wherein each element in the preference attribute vector is used to characterize the target user's preference attributes for different components.

5. The method according to claim 4, characterized in that, The process of determining user behavior characteristics based on the component requirement description information and the historical interaction information of each historical interaction component includes: The user requirement description information is encoded based on a preset encoder to obtain a user requirement description encoding vector. The user demand description encoding vector is used to extract features based on a two-layer fully connected network to obtain user demand description features. For the historical interaction information of each of the historical interaction components, a historical interaction component encoding vector is obtained based on the historical interaction information using a preset encoder; wherein, the historical interaction information includes the interaction frequency information and interaction component attribute information corresponding to the historical interaction component; The historical interaction component features are obtained by extracting features from the encoding vector of the historical interaction component based on a two-layer fully connected network. The average value of the features of each historical interaction component is calculated to obtain the features of the historical interaction components to be spliced. The user demand description features and the historical interaction component features to be assembled are combined to obtain user behavior features.

6. The method according to claim 4, characterized in that, The step of determining component attribute features based on the component dependency network graph and the component attribute information of each component in the component library includes: The component attribute information of each component in the component library is encoded based on a preset encoder to obtain the component attribute encoding vector corresponding to each component. Configure the component attribute encoding vector corresponding to each component as the node parameter information of the corresponding component in the component dependency network graph to obtain the component attribute network graph; The component attribute network graph is input into a pre-trained graph neural network model for feature extraction to obtain the component attribute features corresponding to each component.

7. The method according to claim 3, characterized in that, The step of determining the adaptation attributes of each component relative to the target user based on the component dependency network graph, the preference attributes, and the component adaptability calculation model includes: Based on the component dependency network graph, determine the transition matrix that participates in the iterative operation of the component adaptability operation model; The target user's preference attributes for different components are used as the initialization vector of the component adaptability calculation model; Based on the transition matrix and the initialization vector, the component adaptability operation model is iteratively operated until the offset between the current operation result and the previous operation result of the component adaptability operation model meets the preset condition, and then the current operation result is determined as the target operation result. Based on the target calculation results, the adaptation attributes of each component relative to the target user are determined.

8. A component recommendation device, characterized in that, The device includes: The component recommendation response module is used to respond to the component recommendation operation triggered by the target user, and to obtain the component dependency network graph, the component requirement description information of the target user, and the historical interaction information of each historical interactive component used by the target user in the current editing project; wherein, each node in the component dependency network graph corresponds to a different component in the component library, and the connecting edges between nodes represent the association relationship between components; The component to be filtered module is used to determine multiple components to be filtered from the component library based on the component dependency network graph and the historical interaction information. The adaptation attribute determination module is used to determine the adaptation attribute of each component relative to the target user based on the component dependency network graph, the component attribute information of each component in the component library, the component requirement description information, and the historical interaction information of each historical interaction component. The target component recommendation module is used to filter the plurality of components to be filtered based on each of the adaptation attributes, so as to obtain at least one target recommended component corresponding to the target user.

9. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor to enable the at least one processor to perform the component recommendation method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the component recommendation method of any one of claims 1-7.

Images