A multi-model fusion method based on a graph model association recommendation algorithm

Abstract

The present invention discloses a multi-model fusion method based on a graph model association recommendation algorithm. This method can effectively utilize historical model resources to obtain new business models. The multi-model fusion method includes the following steps: Step A: Obtain multiple historical models with the same object group; Step B: In the historical models, respectively read multiple sequentially connected objects starting from or ending at each object in the object group, and use the multiple repeatedly occurring objects and their object relationships as the relationship chain; Step C: Retrieve multiple historical models with the relationship chain; Step D: If, among the retrieved historical models, the objects on the relationship chain are also connected to other objects in the object group outside the relationship chain, then such an object is an associated object in the historical model, and obtain the number of associated objects in each historical model; Step E: Extract multiple historical models with a relatively large number of associated objects, and integrate these extracted historical models to achieve multi-model fusion to generate a new business model.

Description

Technical Field

[0001] The present invention mainly relates to the technical field of model processing, and particularly relates to a multi-model fusion method based on a graph model association recommendation algorithm. Background Art

[0002] The dynamic ontology modeling tool is an open and autonomous application platform that uses big data analysis technology and knowledge graph technology to meet the needs of data analysis, semantic fusion, business exploration, etc. A graph database is a storage engine specifically used to store and retrieve a huge relational information network. It can efficiently store data as vertices (Nodes) and the relationships between data as edges (Edges), and can also add relevant attributes to vertices or edges, and can achieve efficient full-text retrieval and query of vertices, edges, and attributes. However, after the dynamic ontology modeling tool has been used for a long time, for the hundreds or thousands of business models that have been built, except for a few general business models, targeted models are often discarded after business execution, resulting in waste of historical model resources. And it is very accidental to find valuable business models from the huge historical model database. The problem of how to utilize historical model resources troubles the platform and users. Summary of the Invention

[0003] The technical problem to be solved by the present invention is how to effectively utilize historical model resources to generate new business models.

[0004] To solve the above technical problem, the present invention provides a multi-model fusion method based on a graph model association recommendation algorithm, including the following steps:

[0005] Step A: Obtain multiple historical models with the same object group. The object group refers to: multiple objects connected in a closed loop. The object group includes the respective attributes of these objects and the connection relationships between each object.

[0006] Step B: In the obtained historical models, respectively read multiple sequentially connected objects starting from or ending at each object in the object group, and use multiple objects and their object relationships that repeatedly appear in all the obtained historical models up to a preset degree as a relationship chain.

[0007] Step C: Retrieve multiple historical models having the relationship chain.

[0008] Step D: If, in the retrieved historical models, the objects on the relationship chain are also connected from outside the relationship chain to other objects in the object group, then consider such an object as an associated object in the historical model, and obtain the number of associated objects in each historical model.

[0009] Step E: Extract multiple historical models whose number of associated objects is within a preset range, and integrate these extracted historical models to achieve multi-model fusion to generate a new business model.

[0010] Further, it also includes a display step performed after the step E. According to the user's selection operation on the multiple historical models used to integrate the new business model, display the objects and the relationships between objects of the new business model that are different from the selected historical models.

[0011] Further, in the step A, the multiple objects in the object group being closed-loop connected means that these objects form a cyclic closed loop.

[0012] Further, in the step A, the multiple objects being closed-loop connected means that these objects are all within a closed loop and there is a starting point and / or an ending point of the closed loop among these objects.

[0013] Further, in the step B, recurring up to a preset degree means that the number of repetitions reaches a preset lower limit value and / or the number of repetitions ranks within a preset range among all relationship chains.

[0014] Specifically, in the step B, taking each object in the object group as a starting point / ending point, read N objects connected to it forward / backward respectively, where N is within a preset quantity range. Taking the sequentially connected multiple objects read and the relationships between adjacent objects as an object chain with connection relationships, perform a repeatability check on all object chains read from all historical models. If an object chain recurs up to a preset degree, define this object chain as the relationship chain of these historical models.

[0015] Further, it includes a general model library pre-storing multiple general concept models. The historical models obtained in the step A refer to the historical models whose similarity to each general concept model is not higher than a preset threshold.

[0016] Further, it includes an object element acquisition step performed before the step A to acquire the construction elements of the new business model. The object groups of the multiple historical models obtained in the step A all have objects matching the object elements, and the relationship chains in the step B start or end with the objects matching the object elements.

[0017] Further, the object elements are extracted according to the user's input operation.

[0018] Further, the extraction according to the user's input operation means taking the words input or selected by the user as the object elements.

[0019] Further, the extraction according to the user's input operation means obtaining the natural language input by the user and executing a preset element extraction algorithm, and using the vocabulary extracted by the element extraction algorithm as object elements.

[0020] Advantages of the invention: In the above multi-model fusion method based on the graph model association recommendation algorithm, first, multiple historical models with the same object group are found from the historical model database. Since the object group is formed by the closed-loop connection of multiple objects, and the multiple objects have relationship transmission and multiple-dimensional association relationships among them, the relationship between two historical models with the same object group is relatively close, and there is a basis for mutual association and fusion. Then, among these closely related historical models, multiple objects starting or ending with each object in the object group are found, and the multiple objects that appear repeatedly in multiple historical models are used as relationship chains, that is, branches of the object group. The relationship between historical models with these relationship chains is closer. These historical models are retrieved, and then the other objects in the object group that can be connected from outside the relationship chain on the relationship chain in these historical models are retrieved, that is, the associated objects in the historical model that have an association relationship with both the relationship chain and the object group. The number of associated objects is used to represent the degree of association tightness between each historical model and the object group and its relationship chain branches. The more the number of associated objects, the richer the business objects associated with the object group in the historical model. The new business model generated by fusing these historical models has rich and strongly associated business associated objects, and it is easy to obtain a valuable business model. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 is a flowchart of the multi-model fusion method based on the graph model association recommendation algorithm provided by the present invention;

[0022] Figure 2 is a structural diagram of a historical model in this embodiment;

[0023] Figure 3 is another structural diagram of a historical model in this embodiment;

[0024] Figure 4 is composed of Figure 2 and Figure 3 The business model structure diagram obtained by fusing the historical models of is obtained by the multi-model fusion method based on the graph model association recommendation algorithm provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0025] The following further details the present invention in conjunction with specific embodiments.

[0026] Such as Figure 1The multi - model fusion method based on the graph - model association recommendation algorithm shown is applied to a dynamic ontology modeling tool. The dynamic ontology modeling tool includes a database, a background server, and a terminal device on the user side. The database stores a large number of data tables and a large number of business models that have been constructed in the past and verified through examples. These business models are stored in the historical model database as historical models for users to retrieve and call. Taking the process of a user calling multiple historical models on the dynamic ontology modeling tool for multi - model fusion to generate a new business model as an example, the user submits object elements of the model, such as distribution network equipment, on the terminal device. The server executes the object element acquisition step to obtain the object element "distribution network equipment" as a construction element of the new business model, and then executes the following specific steps of the above - mentioned multi - model fusion method based on the graph - model association recommendation algorithm.

[0027] A. Obtain multiple historical models with the same object group. The object group refers to: multiple objects connected in a closed loop. The object group includes the respective attributes of these objects and the connection relationships between each object. The multiple objects in the object group being connected in a closed loop means that these objects form a circular closed loop, or these objects are all within a closed loop and there is a starting point, an ending point, or both of the closed loop among these objects. Since the object group is formed by multiple objects connected in a closed loop, the multiple objects have relationship transmission and multiple - dimensional association relationships among objects. Therefore, the relationship between two historical models with the same object group is relatively close, having a basis for mutual association and fusion.

[0028] In this embodiment, there can be multiple object groups used to match historical models. The more object groups there are, the fewer historical models are matched. These object groups all have objects that match the object element "distribution network equipment", such as substation equipment belonging to the concept of distribution network equipment and transformers belonging to the concept of substation equipment. Among the historical models obtained after the above - mentioned object element screening, select these historical models with the same object group. The association relationships of these historical models are closer.

[0029] Such as Figure 2 the "distribution network equipment operation model" and Figure 3 the "distribution network equipment value model". Both have the object group of the cycle "distribution network equipment -

trigger

occur

target

occur

associated equipment

occur

[0030] Among them, the historical model database includes a general model library that pre-stores multiple general concept models. The historical models obtained above refer to the historical models whose similarity to each general concept model is not higher than a preset threshold. For example, the number of all objects in the historical model that appear in the same general concept model does not exceed 80%, so as to prevent the general concept model from having a strong correlation with a large number of historical models and interfering with the selection of historical models.

[0031] B. Among these obtained historical models, starting from or ending with each object in the object group, read multiple sequentially connected objects, which repeat to a preset degree in all the obtained historical models. For example, multiple objects and their object relationships whose repetition times reach the preset lower limit value of 10 times and whose repetition times are ranked within the preset range (top 5) in all relationship chains are used as relationship chains. More specifically, the relationship chain starts from or ends with an object that matches the object element, further strengthening the relationship tightness of the historical models.

[0032] Specifically, taking the object group "Distribution network equipment -

Trigger

Occur

Target

Belong to

Own

Execute

Target

Belong to

Have

Asset

[0033] C. Retrieve multiple historical models with relationship chains. For example, the "Distribution network equipment operation model" with the relationship chain: "Distribution network equipment -

Belong to

Own

[0034] D. If, among the retrieved historical models, the objects on the relationship chain are also connected to other objects in the object group outside the relationship chain, then consider this object as an associated object in the historical model, and obtain the number of associated objects in each historical model. For example Figure 3In addition to the relationship chain in C above, the object "unit" in Figure 3 is also connected to the object overhaul in the object group through "unit -

has

asset

occurs

has

asset

[0035] associated object "unit" of is an object with a closer degree of association. After expanding it by 1.5 times, it is then included in the number of associated objects. Figure 2 and Figure 3 After integrating the models of, a new business model that combines the two is obtained. The server provides a function for manual adjustment of the new business model, such as selecting to delete some objects and object relationships. Finally, the adjusted Figure 4 "distribution network equipment overhaul - value change model" is obtained.

[0036] In this embodiment, by executing the multi - model fusion method based on the graph model association recommendation algorithm above, using the object group as the main branch of the association of multiple historical models, and taking the relationship chains in multiple historical models as the branches of their common object group, the historical models with these relationship chains are more closely related. Retrieving these historical models and then retrieving other objects in the object group that can be connected from outside the relationship chain on the relationship chains in these historical models, that is, the associated objects in the historical model that have an association relationship with both the relationship chain and the object group. The number of associated objects represents the degree of closeness of the association between each historical model and the object group and its relationship chain branches. The more the number of associated objects, the richer the business objects associated with the object group in the historical model. The new business model generated by fusing these historical models has rich business - associated objects with strong associations and is easy to obtain a valuable business model.

[0037] Further, it also includes a display step executed after step E. According to the user's selection operation on the multiple historical models used to integrate the new business model, display the objects and the relationships between objects in the new business model that are different from the selected historical models. Utilizing the advantages of the graph model, visually display the richer business objects and object relationships brought by the new business model.

[0038] Further, when obtaining object elements, the object elements are extracted according to the user's input operations. For example, the words input or selected by the user are used as object elements, that is, the user inputs the object element "distribution network equipment" on the terminal device of the dynamic ontology modeling tool; or the server obtains the natural language input by the user and executes a preset element extraction algorithm, and uses the words extracted by the element extraction algorithm as object elements (the element extraction algorithm is implemented by existing vocabulary recognition algorithms and will not be elaborated here).

[0039] The above is only the implementation mode of the present invention, and does not limit the scope of patent protection. Those skilled in the art make non-substantive changes or substitutions based on the present invention, and still fall within the scope of patent protection.

Claims

1. A multi-model fusion method based on a graph model association recommendation algorithm, characterized in that It includes the following steps: Step A: Obtain multiple historical models with the same object group. The object group refers to multiple objects connected in a closed loop, and the object group includes the respective attributes of these objects and the connection relationships between each object; Step B: In the obtained historical models, respectively read multiple sequentially connected objects starting from or ending at each object in the object group, and use multiple objects and their object relationships that repeatedly appear up to a preset degree in all the obtained historical models as the relationship chain; Step C: Retrieve multiple historical models with the said relationship chain; Step D: If, in the retrieved historical models, the objects on the relationship chain are also connected from outside the relationship chain to other objects in the object group, then consider such an object as an associated object in the historical model, and obtain the number of associated objects in each historical model; Step E: Extract multiple historical models whose number of associated objects is within a preset range, and integrate the extracted historical models to achieve multi-model fusion to generate a new business model.

2. The multi-model fusion method based on the graph model association recommendation algorithm according to claim 1, characterized in that: It also includes a display step executed after the said Step E. According to the user's selection operation on the multiple historical models used to integrate the new business model, display the objects and the relationships between objects of the new business model that are different from the selected historical models.

3. The multi-model fusion method based on the graph model association recommendation algorithm according to claim 1, characterized in that: In the said Step A, the multiple objects in the object group being connected in a closed loop means that these objects form a circular closed loop.

4. The multi-model fusion method based on the graph model association recommendation algorithm according to claim 1, characterized in that: In the said Step A, the multiple objects being connected in a closed loop means that these objects are all within a closed loop and there is a starting point and / or an ending point of the closed loop among these objects.

5. The multi-model fusion method based on the graph model association recommendation algorithm according to claim 1, characterized in that: In the said Step B, repeatedly appearing up to a preset degree means that the number of repetitions reaches a preset lower limit value and / or the number of repetitions ranks within a preset range among all the relationship chains.

6. The multi-model fusion method based on the graph model association recommendation algorithm according to claim 1, characterized in that: It includes a general model library pre-storing multiple general concept models. The historical models obtained in the said Step A refer to the historical models whose similarity to each general concept model is not higher than a preset threshold.

7. The multi-model fusion method based on the graph model association recommendation algorithm according to claim 1, characterized in that: It includes an object element acquisition step executed before the said Step A, to acquire the construction elements of the new business model. The object groups of the multiple historical models obtained in the said Step A all have objects matching the object elements, and the relationship chain in the said Step B starts from or ends at the objects matching the object elements.

8. The multi-model fusion method based on the graph model association recommendation algorithm according to claim 7, characterized in that: The object elements are extracted according to the user's input operation.

9. The multi-model fusion method based on the graph model association recommendation algorithm according to claim 8, characterized in that: The extraction according to the user's input operation means using the words input or selected by the user as the object elements.

10. The multi-model fusion method based on the graph model association recommendation algorithm according to claim 8, characterized in that: The extraction according to the user's input operation means obtaining the natural language input by the user, executing a preset element extraction algorithm, and using the words extracted by the element extraction algorithm as the object elements.

Images