A method and device for modeling graph ontology relations for decision tasks

Abstract

The application discloses a kind of graph ontology relationship modeling method and device for decision task, the method includes: to the text material in decision technology field, entity recognition extraction processing is carried out, and the ontology text set is obtained;The ontology text set includes the attribute text set and the relationship text set of each ontology;The attribute text set includes various attribute texts;The relationship text set includes various relationship texts;Ontology relationship modeling processing is carried out to the ontology text set, and the graph ontology relationship prediction model is obtained;The graph ontology relationship prediction model is used to process newly collected ontology attribute text set, and the relationship information of ontology is obtained.

Description

Technical Field

[0001] This invention relates to the fields of knowledge graphs, industrial data processing, and intelligent decision-making technologies, specifically to a graph ontology relationship modeling method and apparatus for decision-making tasks. Background Technology

[0002] In the field of decision-making technology, knowledge graphs, as core tools for carrying domain knowledge and supporting intelligent decision-making, directly determine the reliability of decision results based on the accuracy and completeness of their ontology relationships. With the increasing complexity of decision-making scenarios and the explosive growth of domain data, traditional ontology relationship construction methods have gradually revealed numerous limitations. In existing technologies, the establishment of ontology relationships largely relies on manual annotation or simple rule matching, which not only consumes significant human resources but also struggles to handle complex attribute associations across scenarios and dimensions, resulting in narrow coverage and insufficient flexibility of ontology relationships. Meanwhile, traditional methods, when mapping attributes to relationships, often ignore the nonlinear correlations and dynamic changes between attributes, relying solely on single-dimensional correlation analysis to select attribute factors. This results in weak generalization ability of the constructed relationship models and low prediction accuracy when faced with newly collected ontology data. Furthermore, existing technologies lack refined processing procedures for attribute data, and the presence of noisy data and redundant attributes further affects the accuracy of relationship modeling. Consequently, the decision support capabilities of such models are limited, failing to meet the demand for accurate prediction of ontology relationships in complex decision-making scenarios. On the other hand, the textual data in the decision-making technology field is diverse in type and format, including dictionaries, textbooks, monographs, standards, and specifications. Existing entity recognition and extraction methods are limited and struggle to comprehensively and efficiently extract ontology attribute and relational information from various texts. This results in a weak data foundation for subsequent modeling, further restricting the overall effectiveness of graph ontology relation modeling. Therefore, there is an urgent need for a graph ontology relation modeling method that can efficiently process multi-source textual data, accurately mine the intrinsic connections between attributes and relationships, and possess strong generalization capabilities. This method would address the shortcomings of existing ontology modeling techniques and provide reliable technical support for intelligent decision-making. Summary of the Invention

[0003] This invention primarily addresses the challenge of developing a graph ontology relation modeling method that efficiently processes multi-source text data, accurately mines the intrinsic relationships between attributes and relations, and possesses strong generalization capabilities. This method aims to overcome the shortcomings of existing ontology modeling techniques and provide reliable technical support for intelligent decision-making. This invention discloses a graph ontology relation modeling method and apparatus oriented towards decision-making tasks.

[0004] In a first aspect, this invention discloses a graph ontology relation modeling method for decision-making tasks, comprising: S1, perform entity recognition and extraction processing on textual data in the field of decision-making technology to obtain an ontology text set; the ontology text set includes the attribute text set and relation text set of each ontology; the attribute text set includes various types of attribute text; the relation text set includes various types of relation text. S2, perform ontology relation modeling on the ontology text set to obtain a graph ontology relation prediction model; S3, using the graph ontology relationship prediction model, process the newly collected ontology attribute text set to obtain ontology relationship information; The process of performing ontology relation modeling on the ontology text set to obtain a graph ontology relation prediction model includes: S21, perform text vector conversion processing on the ontology text set to obtain the ontology data set; S22, preprocess the ontology data set to obtain a preprocessed data set; S23, perform ontology relation modeling on the preprocessed data set to obtain a graph ontology relation prediction model.

[0005] The preprocessing of the ontology data set to obtain a preprocessed data set includes: S221, Perform data cleaning processing on the ontology data set to obtain the first dataset; S222, perform data reduction processing on the first dataset to obtain a preprocessed information set; the preprocessed information set includes the attribute data set and relation data set of each ontology.

[0006] The step of performing ontology relation modeling on the preprocessed information set to obtain a graph ontology relation prediction model includes: S31, based on each type of relation value in the relation data set of the preprocessed information set, perform factor filtering on the attribute data set in the preprocessed information set to obtain the attribute data subset corresponding to each type of relation value; S32, For each type of relation value, perform regression analysis and modeling with the corresponding attribute data subset to obtain the corresponding ontology relation prediction model; S33. Using the ontology relation prediction model of relation values ​​of all classes, a graph ontology relation prediction model is constructed.

[0007] The step involves performing factor filtering on the attribute data set in the preprocessed information set for each type of relation value in the relation data set, to obtain a subset of attribute data corresponding to each type of relation value, including: S311, based on each type of relation value in the relation data set of the preprocessed information set, perform preliminary screening on the attribute data set in the preprocessed information set to obtain the first related attribute data set for each type of relation value; S312, perform association pattern screening on the first related attribute data set of each type of relation value to obtain the attribute data subset corresponding to the type of relation value.

[0008] The method involves performing preliminary filtering on the attribute data set in the preprocessed information set for each type of relation value in the relation data set, to obtain a first related attribute data set for each type of relation value, including: S3111, based on each type of relation value in the relation data set of the preprocessed information set, perform relevant modeling processing on the attribute data set in the preprocessed information set to obtain the relevant model of the type of relation value; S3112, for each type of attribute variable in the attribute data set of the preprocessed information set, the attribute variable is discretized at multiple levels within its value range to form N value intervals; M sample points are randomly selected in each value interval to obtain a sample point set, the number of sample points in the sample point set being M; the value intervals of each type of attribute variable are combined to obtain a value interval combination set; the value interval combination set includes several value interval combinations; the value interval combination includes the value interval of each type of attribute variable; S3113, for each value interval combination, use all the sample points in the value intervals it contains to combine them to obtain the sample point combination of the value interval combination; each sample point in the sample point combination corresponds to an attribute data in the attribute data set. S3114, using the relevant model of each type of relation value, calculate and process the sample point combination for each value interval combination to obtain the predicted output of the corresponding relation value; S3115, For each type of relation value, calculate the total output variance of the predicted output for all sample point combinations under all value intervals of the attribute variable. Then, the normalized contribution rate of each value interval combination is calculated. S3116, Determine whether the normalized contribution rate of each value interval combination is greater than a preset threshold. The first discrimination result is obtained; if the first discrimination result is yes, the corresponding attribute data of all sample points of the value interval combination in the attribute data set in the preprocessed information set is added to the first related attribute data set.

[0009] The formula for calculating the normalized contribution rate is: in, The normalized contribution rate for the i-th value interval combination. Let be the variance of the predicted output for all sample point combinations of the i-th value interval combination. This represents the total output variance.

[0010] A second aspect of this invention discloses a graph ontology relation modeling apparatus for decision-making tasks, the apparatus comprising: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the graph ontology relation modeling method for decision-making tasks.

[0011] In a third aspect, the present invention discloses a computer-readable storage medium storing computer instructions, which, when invoked by a computer, are used to execute the graph ontology relation modeling method for decision-oriented tasks.

[0012] In a fourth aspect of this invention, an information data processing terminal is disclosed, which is used to implement the graph ontology relationship modeling method for decision-making tasks.

[0013] The beneficial effects of this invention are as follows: This invention integrates multiple entity recognition and extraction methods based on rules, dictionaries, and online knowledge bases, enabling comprehensive and efficient extraction of attribute and relational texts of ontology from various textual materials in the field of decision technology. This constructs a complete ontology text set, laying a solid data foundation for subsequent ontology relation modeling and effectively solving the problems of incomplete data extraction and low efficiency caused by the single nature of traditional entity recognition methods.

[0014] This invention, based on a filtered subset of attribute data, constructs a delay vector, performs nonlinear dimensionality reduction, and establishes a dynamic response model between relation values ​​and a two-dimensional latent variable sequence. It fully considers the dynamic change characteristics and nonlinear correlation of attribute data. The constructed ontology relation prediction model can accurately capture the complex mapping relationship between attributes and relations, significantly improving the accuracy of ontology relation prediction. Compared with traditional linear modeling methods, it has stronger adaptability and predictive ability. This invention integrates ontology relation prediction models corresponding to all types of relation values ​​to construct a graph ontology relation prediction model. This model can simultaneously handle the prediction needs of multiple types of ontology relations without having to build a separate model for each type of relation, thus significantly improving modeling efficiency. Furthermore, this model can adapt to newly collected ontology attribute text sets and can quickly predict relations through entity recognition and extraction. It has good generalization ability and scalability, and can meet the ontology relation modeling needs in different decision-making scenarios.

[0015] This invention constructs a relevant model through polynomial function fitting and performs preliminary attribute screening by combining normalized contribution rate analysis. This can quantify the influence of attribute variables on relation values, making the attribute factor screening process more scientific and objective. It avoids the bias caused by subjective experience judgment in traditional methods and further improves the reliability and rationality of ontology relation modeling.

[0016] The dynamic response model of this invention, by introducing parameters such as the steepness of the control inhibition curve and the critical mass transfer threshold, can flexibly adapt to the dynamic change patterns of different attribute data sequences, accurately characterize the nonlinear characteristics of the relationship value as the attribute data changes, and is more in line with the complex characteristics of ontological relationships in the field of decision technology, making the prediction results more consistent with actual application scenarios. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the implementation of the method of the present invention. Detailed Implementation

[0018] To better understand the content of this invention, an embodiment is provided here.

[0019] Figure 1 This is a flowchart illustrating the implementation of the method of the present invention.

[0020] In a first aspect, this invention discloses a graph ontology relation modeling method for decision-making tasks, comprising: S1, perform entity recognition and extraction processing on textual data in the field of decision-making technology to obtain an ontology text set; the ontology text set includes the attribute text set and relation text set of each ontology; the attribute text set includes various types of attribute text; the relation text set includes various types of relation text. S2, perform ontology relation modeling on the ontology text set to obtain a graph ontology relation prediction model; S3, using the graph ontology relationship prediction model, process the newly collected ontology attribute text set to obtain ontology relationship information; The entity recognition and extraction processing method includes rule-based entity recognition, dictionary-based entity recognition, and online knowledge base methods.

[0021] The newly acquired set of ontology attribute texts can be used to perform entity recognition and extraction processing on the newly acquired text data, and then the attribute text set can be obtained from the extraction results.

[0022] Textual materials in the field of decision-making technology include existing dictionaries, classic textbooks, monographs, standards and specifications in decision-related fields.

[0023] The attribute text refers to the text of various attribute information of the ontology. For example, the attribute text of a geographic name ontology includes geographic coordinates, value range, spatial precision, etc. The relationship text includes various relationships between the ontology and other ontologies, such as projection method, adjacency relationship, overlap area, etc. The ontology can be an entity in the knowledge graph domain.

[0024] The process of performing ontology relation modeling on the ontology text set to obtain a graph ontology relation prediction model includes: S21, perform text vector conversion processing on the ontology text set to obtain the ontology data set; S22, preprocess the ontology data set to obtain a preprocessed data set; S23, perform ontology relation modeling on the preprocessed data set to obtain a graph ontology relation prediction model.

[0025] The text vector conversion process can be implemented using the word2vec command or the text vector conversion command in Python.

[0026] The preprocessing of the ontology data set to obtain a preprocessed data set includes: The ontology dataset is cleaned to obtain the first dataset; The first dataset is subjected to data reduction processing to obtain a preprocessed information set; the preprocessed information set includes an attribute data set and a relation data set for each ontology; the relation data set includes relation data, which includes relation values ​​of several types of relations; the attribute data set includes attribute data sequences of several types of attribute variables; the attribute data sequence includes several attribute data. The data cleaning process includes smoothing noisy data and smoothing or deleting outliers; The data reduction process can be performed using dimensional reduction or numerical reduction.

[0027] The step of performing ontology relation modeling on the preprocessed information set to obtain a graph ontology relation prediction model includes: S31, based on each type of relation value in the relation data set of the preprocessed information set, perform factor filtering on the attribute data set in the preprocessed information set to obtain the attribute data subset corresponding to each type of relation value; S32, For each type of relation value, perform regression analysis and modeling with the corresponding attribute data subset to obtain the corresponding ontology relation prediction model; S33. Using the ontology relation prediction model of relation values ​​of all classes, a graph ontology relation prediction model is constructed.

[0028] The step involves performing factor filtering on the attribute data set in the preprocessed information set for each type of relation value in the relation data set, to obtain a subset of attribute data corresponding to each type of relation value, including: S311, based on each type of relation value in the relation data set of the preprocessed information set, perform preliminary screening on the attribute data set in the preprocessed information set to obtain the first related attribute data set for each type of relation value; S312, perform association pattern screening on the first related attribute data set of each type of relation value to obtain the attribute data subset corresponding to the type of relation value.

[0029] The method involves performing preliminary filtering on the attribute data set in the preprocessed information set for each type of relation value in the relation data set, to obtain a first related attribute data set for each type of relation value, including: S3111, based on each type of relation value in the relation data set of the preprocessed information set, perform relevant modeling processing on the attribute data set in the preprocessed information set to obtain the relevant model of the type of relation value; S3112, for each type of attribute variable in the attribute data set of the preprocessed information set, the attribute variable is discretized at multiple levels within its value range to form N value intervals (N≥5); M sample points are randomly selected from each value interval to obtain a sample point set, the number of sample points in the sample point set being M (M≥10); the value intervals of each type of attribute variable are combined to obtain a value interval combination set; the value interval combination set includes several value interval combinations; the value interval combination includes the value interval of each type of attribute variable; S3113, for each value interval combination, use all the sample points in the value intervals it contains to combine them to obtain the sample point combination of the value interval combination; each sample point in the sample point combination corresponds to an attribute data in the attribute data set. The combination of value ranges is obtained by selecting one value range from each type of attribute variable and combining them. The step of randomly selecting M sample points within each value interval involves randomly selecting all values ​​of the attribute variable within each value interval to obtain M sample points, and using these M sample points to obtain a sample point set; all values ​​of the attribute variable within each value interval are all values ​​of the attribute variable in the attribute data set.

[0030] S3114, using the relevant model of each type of relation value, calculate and process the sample point combination for each value interval combination to obtain the predicted output of the corresponding relation value; S3115, For each type of relation value, calculate the total output variance of the predicted output for all sample point combinations under all value intervals of the attribute variable. Then, the normalized contribution rate of each value interval combination is calculated. S3116, Determine whether the normalized contribution rate of each value interval combination is greater than a preset threshold. The first discrimination result is obtained; if the first discrimination result is yes, the corresponding attribute data of all sample points of the value interval combination in the attribute data set in the preprocessed information set is added to the first related attribute data set. The formula for calculating the normalized contribution rate is: in, The normalized contribution rate for the i-th value interval combination. Let be the variance of the predicted output for all sample point combinations of the i-th value interval combination. This represents the total output variance.

[0031] The normalized contribution rate calculation expression quantifies the ratio of the output variance of a single value interval combination to the total output variance. It can objectively reflect the degree of influence of each value interval combination on the predicted output of the relation value, providing a scientific and quantitative basis for the selection of value interval combinations. This avoids the bias caused by subjective experience selection, effectively retains the attribute data corresponding to the value interval combinations that have a significant contribution to relation prediction, improves the pertinence and reliability of the first relevant attribute data set, and lays the foundation for the accurate selection of subsequent attribute factors.

[0032] In the process of ontology relation modeling, this invention first performs data cleaning and data reduction preprocessing on the ontology dataset. By smoothing noisy data, deleting outliers, and performing dimensional or numerical reduction, redundant information and abnormal data are eliminated, significantly improving the quality of the modeling data. This ensures the accuracy of the subsequent relation prediction model and avoids model prediction bias caused by data quality issues. This invention employs a two-stage factor screening process for each type of relation value, involving preliminary screening and refined screening based on association patterns. First, a first set of related attribute data is obtained based on correlation modeling and multi-level discretization analysis. Then, refined screening is performed by calculating the conditional co-occurrence strength index and screening the total number of attribute data. This process can accurately identify subsets of attribute data that are strongly correlated with various relation values, effectively eliminating interference from irrelevant and weakly correlated attributes, and making the attribute factors relied upon for subsequent modeling more targeted.

[0033] The process of performing correlation modeling on the attribute data set in the preprocessed information set for each type of relation value in the relation data set is as follows: Each type of relation value in the relation data set is the dependent variable, and each type of attribute variable in the attribute data set is the multivariate independent variable. A multinomial function is fitted to obtain the correlation model for the relation values, and its expression is: In the formula: Y is the predicted output of the relation value. Let be the values ​​of the j-th and k-th attribute variables, respectively. For constant terms, ε represents the coefficients of the linear, quadratic, and interaction terms to be estimated, respectively; ε is the model residual term; and P is the total number of attribute variables.

[0034] The polynomial function fitting expression of the correlation model, by integrating linear, quadratic, and interaction terms, comprehensively captures the multivariate correlation characteristics between attribute variables and relation values. It can not only reflect the linear influence of a single attribute variable on relation values, but also accurately characterize the nonlinear effect of the attribute variable itself and the interaction between different attribute variables. Compared with traditional single-dimensional correlation analysis, it is more in line with the complex characteristics of ontological relationships in the field of decision technology, and provides more comprehensive and accurate correlation modeling support for the preliminary screening of subsequent attribute data, ensuring that the screened attribute data can cover the key factors affecting relation values.

[0035] The first set of related attribute data is constructed through initialization before executing S3111. The value satisfies 0.05 ≤ T 1≤0.15.

[0036] The first related attribute data set for each type of relation value is subjected to association pattern screening to obtain the attribute data subset corresponding to the type of relation value, including: S3121, Based on the value distribution of each type of relation, the value distribution of each type of relation is divided into low correlation (<30% maximum value), medium correlation (30%-70%) and high correlation (>70%). S3122, Statistically obtain the attribute data of each attribute variable in the first relevant attribute data set for each type of relation value. C Relationship value with each high correlation H Frequency of occurrence Statistical analysis yields data for each attribute. C Frequency of occurrence ; S3123, based on frequency and The conditional co-occurrence intensity index of each attribute data is calculated. ; The calculation expression for the conditional co-occurrence intensity index is as follows: , in, The i-th relation value with high correlation The frequency of occurrence, where V is the total number of highly correlated relation values ​​in a class of relation values. For attribute data C Relationship value with each high correlation Frequency of simultaneous occurrence.

[0037] S3124, Delete the attribute data whose conditional co-occurrence intensity index is less than the preset intensity threshold from the first related attribute data set; S3125, Calculate the total number of attribute data for each type of attribute variable in the first related attribute data set; S3126, delete all attribute data of attribute variables whose total number of attribute data is less than the preset total number of data from the first related attribute data set to obtain the attribute data subset corresponding to the class relationship value.

[0038] The calculation expression of the conditional co-occurrence strength index comprehensively quantifies the co-occurrence strength between attribute data and highly correlated values ​​by combining the co-occurrence frequency of attribute data and highly correlated values, the frequency of attribute data occurring alone, and the frequency of highly correlated values ​​occurring on their own. At the same time, the sine function and logarithmic operation are used to enhance the distinguishability of strong co-occurrence relationships. It can effectively identify attribute data with stable and strong correlations with highly correlated values, and eliminate attribute data with weak correlations or occasional co-occurrences. This further improves the accuracy of the attribute data subset, ensures that the attribute factors on which subsequent modeling depends have a strong correlation with the target relational value, and reduces the interference of redundant attributes on model performance.

[0039] The process of performing regression analysis and modeling on each type of relation value and its corresponding attribute data subset to obtain the corresponding ontology relation prediction model includes: S321, for the attribute data sequence of each type of attribute variable in the attribute data subset corresponding to each type of relation value, construct a delay vector, wherein the expression of the t-th element of the delay vector is: in, The time delay index is determined by the minimum index value of the element whose first zero-crossing point is the autocorrelation function of the attribute data sequence of all class attribute variables. Let t be the t-th element of the attribute data sequence of a class of attribute variables; S322, using all the delay vectors corresponding to each type of relation value, construct a set of delay vectors; S323, Perform nonlinear dimensionality reduction on the delay vector set to obtain a two-dimensional latent variable sequence. The expression for the t-th element of the two-dimensional latent variable sequence is: , Each element is obtained by weighted summation of the elements of the delay vector; S324, Establish a dynamic response model between each type of relation value and the two-dimensional latent variable sequence: in, Let t be the t-th value of a relation, corresponding to the t-th element of the attribute data sequence for each type of attribute variable. Let be the weight coefficients to be solved. The steepness of the control suppression curve to be solved. This is the critical threshold for mass transfer.

[0040] S325, Solve for the unsolved variables of the dynamic response model to obtain... and The value of is substituted into the dynamic response model to obtain the ontology relationship prediction model corresponding to the class relationship value.

[0041] The solution to the unsolved variables of the dynamic response model can be achieved using numerical iterative solution methods or particle filtering methods, etc.

[0042] The expression for the delay vector, by integrating the values ​​at different time points in the attribute data sequence, fully preserves the temporal dimension correlation information of the attribute data. It can capture the dynamic characteristics of attribute data changing over time, avoiding the limitation of traditional static data modeling ignoring temporal correlation. It provides comprehensive data support containing historical change information for subsequent nonlinear dimensionality reduction and dynamic response models, enabling the model to better adapt to the ontology relationship prediction needs under dynamic changes in attribute data and improve the model's ability to capture dynamic relationships.

[0043] The expression for the two-dimensional latent variable sequence achieves nonlinear dimensionality reduction by weighted summation of the elements of the delay vector. While significantly simplifying the data dimension and reducing the computational complexity of the model, it can retain the key feature information in the delay vector to the greatest extent and avoid the loss of core correlation information during the dimensionality reduction process. This not only reduces the computational burden of the subsequent dynamic response model, but also provides the model with concise and effective input data, achieving a balance between model computational efficiency and prediction accuracy, and is suitable for relational modeling scenarios of large-scale ontology data.

[0044] The dynamic response model integrates linear terms with suppression curve terms based on critical thresholds, enabling it to flexibly characterize the nonlinear dynamics of relation values ​​as two-dimensional latent variables change. It can accurately capture the impact of linear changes in latent variables on relation values ​​and effectively adapt to the abrupt changes in relation values ​​under critical conditions. This fully aligns with the complex and variable nature of ontology relations in the field of decision-making technology. Compared to traditional linear regression models, it can more accurately reflect the dynamic trends of relation values, significantly improving the accuracy of ontology relation prediction and its adaptability to complex scenarios. It ensures that the model can output reliable relation prediction results under different attribute data distributions.

[0045] In all embodiments of the present invention, the variables involved in all computational expressions or mathematical functions have been dimensionlessized before computation.

[0046] In all embodiments of the present invention, the values ​​of the independent variables in the input of all computational expressions or mathematical functions meet the reasonable requirements of the input range of the computational expressions or mathematical functions, and can ensure that the computational expressions or mathematical functions can be calculated smoothly without violating physical laws or mathematical rules.

[0047] A second aspect of this invention discloses a graph ontology relation modeling apparatus for decision-making tasks, the apparatus comprising: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the graph ontology relation modeling method for decision-making tasks.

[0048] In a third aspect, the present invention discloses a computer-readable storage medium storing computer instructions, which, when invoked by a computer, are used to execute the graph ontology relation modeling method for decision-oriented tasks.

[0049] In a fourth aspect of this invention, an information data processing terminal is disclosed, which is used to implement the graph ontology relationship modeling method for decision-making tasks.

[0050] The above description is merely an embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.

Claims

1. A graph ontology relation modeling method for decision-making tasks, characterized in that, include: S1, perform entity recognition and extraction processing on textual data in the field of decision-making technology to obtain an ontology text set; the ontology text set includes the attribute text set and relation text set of each ontology; the attribute text set includes various types of attribute text; the relation text set includes various types of relation text. S2, perform ontology relation modeling on the ontology text set to obtain a graph ontology relation prediction model, including: S21, perform text vector conversion processing on the ontology text set to obtain the ontology data set; S22, preprocess the ontology data set to obtain a preprocessed information set; S31, based on each type of relation value in the relation data set of the preprocessed information set, perform factor filtering on the attribute data set in the preprocessed information set to obtain a subset of attribute data corresponding to each type of relation value, including: S3111, based on each type of relation value in the relation data set of the preprocessed information set, perform relevant modeling processing on the attribute data set in the preprocessed information set to obtain the relevant model of the type of relation value; S3112, for each type of attribute variable in the attribute data set of the preprocessed information set, the attribute variable is discretized at multiple levels within its value range to form N value intervals; M sample points are randomly selected in each value interval to obtain a sample point set, the number of sample points in the sample point set being M; the value intervals of each type of attribute variable are combined to obtain a value interval combination set; the value interval combination set includes several value interval combinations; the value interval combination includes the value interval of each type of attribute variable; S3113, for each value interval combination, use all the sample points in the value intervals it contains to combine them to obtain the sample point combination of the value interval combination; each sample point in the sample point combination corresponds to an attribute data in the attribute data set. S3114, using the relevant model of each type of relation value, calculate and process the sample point combination for each value interval combination to obtain the predicted output of the corresponding relation value; S3115, For each type of relation value, calculate the total output variance of the predicted output for all sample point combinations under all value intervals of the attribute variable. Then, the normalized contribution rate of each value interval combination is calculated. S3116, Determine whether the normalized contribution rate of each value interval combination is greater than a preset threshold. The first discrimination result is obtained; if the first discrimination result is yes, the corresponding attribute data of all sample points of the value interval combination in the attribute data set in the preprocessed information set is added to the first related attribute data set. S312, perform association pattern screening on the first related attribute data set of each type of relation value to obtain the attribute data subset corresponding to the type of relation value; S32, For each type of relation value, perform regression analysis and modeling with the corresponding attribute data subset to obtain the corresponding ontology relation prediction model; S33, using the ontology relation prediction model of relation values ​​of all classes, a graph ontology relation prediction model is constructed; S3. Using the graph ontology relationship prediction model, the newly collected ontology attribute text set is processed to obtain ontology relationship information.

2. The graph ontology relation modeling method for decision-making tasks as described in claim 1, characterized in that, The preprocessing of the ontology data set to obtain a preprocessed information set includes: S221, Perform data cleaning processing on the ontology data set to obtain the first dataset; S222, perform data reduction processing on the first dataset to obtain a preprocessed information set; the preprocessed information set includes the attribute data set and relation data set of each ontology.

3. The graph ontology relation modeling method for decision-making tasks as described in claim 1, characterized in that, The formula for calculating the normalized contribution rate is: in, The normalized contribution rate for the i-th value interval combination. Let be the variance of the predicted output for all sample point combinations of the i-th value interval combination. This represents the total output variance.

4. A graph ontology relation modeling device for decision-making tasks, characterized in that, The device includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the graph ontology relation modeling method for decision-making tasks as described in any one of claims 1 to 3.

5. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions, which, when invoked by a computer, are used to execute the graph ontology relation modeling method for decision-oriented tasks as described in any one of claims 1 to 3.

6. An information data processing terminal, characterized in that, The information data processing terminal is used to implement the graph ontology relationship modeling method for decision-making tasks as described in any one of claims 1 to 3.

Images