CN114880457A - Training method of process recommendation model, process recommendation method and electronic equipment - Google Patents

Training method of process recommendation model, process recommendation method and electronic equipment Download PDF

Info

Publication number
CN114880457A
CN114880457A CN202210501295.5A CN202210501295A CN114880457A CN 114880457 A CN114880457 A CN 114880457A CN 202210501295 A CN202210501295 A CN 202210501295A CN 114880457 A CN114880457 A CN 114880457A
Authority
CN
China
Prior art keywords
similarity
graph
sample processing
target
pair
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210501295.5A
Other languages
Chinese (zh)
Inventor
裘超超
李斌
严翼飞
顾峤
程少杰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shanghai Youji Industrial Software Co ltd
Original Assignee
Shanghai Youji Industrial Software Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shanghai Youji Industrial Software Co ltd filed Critical Shanghai Youji Industrial Software Co ltd
Priority to CN202210501295.5A priority Critical patent/CN114880457A/en
Publication of CN114880457A publication Critical patent/CN114880457A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/335—Filtering based on additional data, e.g. user or group profiles
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/332—Query formulation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/338—Presentation of query results
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
    • G06F16/367—Ontology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/30—Semantic analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Databases & Information Systems (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Health & Medical Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Mathematical Physics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Biophysics (AREA)
  • Software Systems (AREA)
  • Biomedical Technology (AREA)
  • Animal Behavior & Ethology (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The invention relates to the technical field of process recommendation, in particular to a training method and a process recommendation method of a process recommendation model and electronic equipment, wherein the training method comprises the steps of obtaining a sample processing characteristic subgraph pair and a target similarity between the sample processing characteristic subgraph pair, wherein the target similarity is determined according to a similarity measurement mode corresponding to an attribute type of each node in the sample processing characteristic subgraph pair, and the attribute type comprises a quantitative attribute and a semantic attribute; inputting the sample processing characteristic subgraph pair into a preset process recommendation model to obtain prediction similarity; and adjusting parameters of the preset process recommendation model based on the difference between the prediction similarity and the target similarity to determine the trained target process recommendation model. The proposed process recommendation model recommends for the process knowledge graph, and accuracy and recommendation efficiency of the target process recommendation model are improved by converting graph structure data of different node attribute types into a vector form and performing calculation among vectors.

Description

Training method of process recommendation model, process recommendation method and electronic equipment
Technical Field
The invention relates to the technical field of process recommendation, in particular to a process recommendation model training method, a process recommendation method and electronic equipment.
Background
With the continued development and widespread use of CAD/CAM systems, a large number of digitized three-dimensional CAD models and associated process data/knowledge are continually generated and stored in the enterprise's data/knowledge base. For enterprises, manufactured products are not completely independent, and even if the enterprises are updated, the structural design, the numerical control process and other aspects of new and old products still have certain similarity and inheritance. Based on this, various process recommendations have been made.
In the process recommendation, case-based reasoning (CBR) is an important branch of artificial intelligence, and mainly carries out analogy reasoning based on old cases or experiences to simulate the thinking and the method of solving problems of human beings. When the process recommendation is carried out, the CBR searches the most similar examples in the example library, then selects to directly adopt the example results according to a certain rule or adopts the results after modifying the results, and stores the examples to realize the dynamic learning of the example library. At present, the most common retrieval algorithm in the CBR is a nearest neighbor method, that is, similarity between a current instance and an instance in an instance library is calculated according to a certain rule, and then a retrieval result is determined according to the size of the similarity. The biggest difficulty of the method is that the difference distance between different examples is reasonably evaluated, and particularly, for semantic text data, the similarity measurement is difficult. Further, the nearest neighbor method is essentially a brute force search method, which has a significantly reduced retrieval speed as the instance base is enlarged.
Disclosure of Invention
In view of this, embodiments of the present invention provide a method for training a process recommendation model, a method for recommending a process, and an electronic device, so as to solve the problem of low efficiency of process recommendation.
According to a first aspect, an embodiment of the present invention provides a method for training a process recommendation model, including:
acquiring a sample processing characteristic subgraph pair and target similarity between the sample processing characteristic subgraph pair, wherein the target similarity is determined according to a similarity measurement mode corresponding to an attribute type of each node in the sample processing characteristic subgraph pair, and the attribute type comprises a quantitative attribute and a semantic attribute;
inputting the sample processing characteristic subgraph pair into a preset process recommendation model to obtain prediction similarity;
and adjusting parameters of the preset process recommendation model based on the difference between the predicted similarity and the target similarity so as to determine a trained target process recommendation model.
The training method of the process recommendation model provided by the embodiment of the invention realizes instance description of process data by using a knowledge graph method, is used for actual retrieval based on a graph similarity measurement algorithm, and calculates the similarity by combining a similarity measurement mode corresponding to the attribute type of each node in a sample processing characteristic subgraph pair in the actual retrieval process, namely, by converting graph structure data of different node attribute types into a vector form and calculating between vectors, the calculation of the similarity between the nodes of various types is ensured, the accuracy and the recommendation efficiency of the trained target process recommendation model are improved, and correspondingly, the efficiency of subsequently recommending processes by using the target process recommendation model is improved.
With reference to the first aspect, in a first implementation manner of the first aspect, acquiring a target similarity between the sample processing feature sub-graph pairs includes:
determining the corresponding attribute type based on the data type of each node in the sample processing characteristic subgraph pair;
determining the similarity between corresponding nodes in the sample processing characteristic subgraph pair by using a similarity measurement mode corresponding to the attribute type;
and fusing the similarity between corresponding nodes in the sample processing characteristic subgraph pair to determine the target similarity.
According to the training method of the process recommendation model, the corresponding attribute types are obtained through the data types of all the nodes in the sample processing characteristic subgraph, so that an accurate similarity measurement mode is determined, the similarities among the nodes are fused, and the accuracy of the determined target similarity is improved.
With reference to the first implementation manner of the first aspect, in a second implementation manner of the first aspect, the determining, by using a similarity measure manner corresponding to the attribute type, a similarity between corresponding nodes in the sample processing feature subgraph pair includes:
when the attribute type is the quantitative attribute, performing similarity calculation based on the numerical values of the corresponding nodes in the sample processing characteristic subgraph pair to obtain the similarity;
and when the attribute type is the semantic attribute, performing similarity calculation based on the editing distance between the character strings of the corresponding nodes in the sample processing characteristic subgraph pair to obtain the similarity.
According to the training method of the process recommendation model provided by the embodiment of the invention, when the attribute type is the quantitative attribute, corresponding data are directly utilized to carry out similarity calculation so as to reduce data processing amount brought by data conversion; when the attribute type is the semantic attribute, the editing distance between the character strings is used for carrying out similarity calculation, and the editing distance is used for measuring the similarity between different words and describing the similarity between the two character strings, so that the similarity between nodes of the semantic attribute is described by using the editing distance, and the accuracy of the similarity calculation of the semantic attribute is improved.
With reference to the first implementation manner of the first aspect, in a third implementation manner of the first aspect, the fusing the similarities between the corresponding nodes in the sample processing feature subgraph pair to determine the target similarity includes:
acquiring weights corresponding to the similarity;
and performing weighted calculation based on the weight and the corresponding similarity to determine the target similarity.
According to the training method of the process recommendation model provided by the embodiment of the invention, the similarity degrees are fused in a weighting mode, the target similarity degree can be determined through reduced calculated amount, and the data processing efficiency is improved.
With reference to the first aspect or any one of the first to third embodiments, in a fourth embodiment of the first aspect, the inputting the sample processing feature subgraph pair into a preset process recommendation model to obtain a prediction similarity includes:
vectorizing the sample processing feature sub-graph pair by using a vector module in the preset process recommendation model to obtain a sample processing feature vector pair with the same dimension;
and performing similarity calculation on the sample processing feature vector pair by utilizing a similarity module in the preset process recommendation model and a similarity measurement mode corresponding to the attribute type of each node in the sample processing feature subgraph to determine the prediction similarity.
According to the training method of the process recommendation model provided by the embodiment of the invention, before the similarity module is processed, the input sample processing feature sub-graph pairs are processed into the sample processing feature vector pairs with the same dimensionality, so that the subsequent calculation of the similarity can be ensured.
With reference to the fourth implementation manner of the first aspect, in the fifth implementation manner of the first aspect, the vectorizing the sample processing feature sub-graph pair by using the vector module in the preset process recommendation model to obtain a sample processing feature vector pair with the same dimension includes:
vectorizing the nodes of the sample processing characteristic sub-graph pair by using a graph convolution network module in the vector module to obtain a node vector pair;
and carrying out graph vectorization on the node vector pair by using an attention network module in the vector module to obtain the sample processing characteristic sub-graph vector pair with the same dimension.
The training method of the process recommendation model provided by the embodiment of the invention vectorizes the nodes of the sample processing characteristic subgraph pair, namely, embeds the nodes; and then, vectorization at the graph level, namely graph embedding is carried out to obtain a sample processing feature vector pair, and the obtained sample processing feature sub-graph pair is sequentially vectorized according to the hierarchical relationship, so that the obtained sample processing feature relatively retains the information of each node and integrates the integral information of the sample processing feature sub-graph, and the accuracy of the obtained sample processing feature vector pair is improved.
According to a second aspect, an embodiment of the present invention further provides a process recommendation method, including:
acquiring a processing characteristic to be recommended;
constructing a processing feature sub-graph to be recommended based on the relation between the processing features;
inputting the processing feature sub-graph to be recommended and each processing feature sub-graph in a process knowledge graph into a target process recommendation model to obtain at least one target processing feature sub-graph with the highest similarity to the processing feature sub-graph to be recommended in the process knowledge graph, wherein the process knowledge graph comprises the processing feature sub-graph and the processing process sub-graph, the processing feature sub-graph and the processing process sub-graph have a corresponding relation, and the target process recommendation model is obtained by training according to the training method of the process recommendation model in the first aspect or any one of the embodiments of the first aspect of the invention;
and searching in the process knowledge graph based on the target processing characteristic subgraph, and determining a target processing process subgraph corresponding to the target processing characteristic subgraph to determine a recommended processing process.
According to the process recommendation method provided by the embodiment of the invention, the target process recommendation model obtained by training is used for determining at least one target processing characteristic subgraph, the target process recommendation model is processed based on graph data, and the similarity calculation is distinguished and calculated based on the attribute type of the node, so that the determination efficiency of the target processing characteristic subgraph is improved, the process knowledge graph is retrieved on the basis, and the retrieval efficiency of the target processing characteristic subgraph is ensured.
With reference to the second aspect, in a first embodiment of the second aspect, the constructing a processing feature subgraph to be recommended based on the relationship between the processing features includes:
determining a father node corresponding to the machining features based on the relation among the machining features;
and constructing the processing feature subgraph to be recommended based on the father node and the processing feature.
According to the process recommendation method provided by the embodiment of the invention, the parent node is determined by utilizing the relation among the processing characteristics, and the child node is determined on the basis, so that the accuracy of the processing characteristic subgraph to be recommended is improved.
According to a third aspect, an embodiment of the present invention provides an electronic device, including: a memory and a processor, the memory and the processor being communicatively coupled to each other, the memory having stored therein computer instructions, and the processor executing the computer instructions to perform the method for training a process recommendation model according to the first aspect or any one of the embodiments of the first aspect, or to perform the method for process recommendation according to any one of the embodiments of the second aspect or the second aspect.
According to a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute the method for training a process recommendation model according to the first aspect or any one of the embodiments of the first aspect, or execute the method for process recommendation according to any one of the embodiments of the second aspect or the second aspect.
It should be noted that, for the corresponding effects of the electronic device or the computer-readable storage medium provided in the embodiment of the present invention, please refer to the description of the corresponding beneficial effects of the process recommendation model training method or the process recommendation method in the foregoing, which is not described herein again.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
FIG. 1 is a flow chart of a method of training a process recommendation model according to an embodiment of the invention;
FIG. 2 is a schematic illustration of a sample processing feature sub-graph according to an embodiment of the invention;
FIG. 3 is a flow chart of a method of training a process recommendation model according to an embodiment of the present invention;
FIG. 4 is a flow chart of a method of training a process recommendation model according to an embodiment of the present invention;
FIG. 5 is a schematic diagram of a training of a process recommendation model according to an embodiment of the invention;
FIG. 6 is a schematic diagram of a training of a process recommendation model according to an embodiment of the invention;
FIG. 7 is a schematic illustration of a process recommendation method according to an embodiment of the present invention;
FIG. 8 is a schematic illustration of a process knowledge map according to an embodiment of the invention;
FIG. 9 is a block diagram of an apparatus for training a process recommendation model according to an embodiment of the present invention;
FIG. 10 is a block diagram of a process recommendation device according to an embodiment of the present invention;
fig. 11 is a schematic diagram of a hardware structure of an electronic device according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The training method and the process recommendation method of the process recommendation model provided by the embodiment of the invention introduce a graph similarity measurement method on the basis of example reasoning, and are used for solving the problems existing in the prior process recommendation by adopting a CBR method, namely: knowledge representation cannot be effectively and uniformly represented, and the expandability and the flexibility are lacked, so that the process knowledge is difficult to reuse and share; the existing example retrieval method is low in efficiency, time-consuming in recommending process knowledge, and based on the problems of a nearest neighbor method in the aspects of text similarity measurement, measurement efficiency and the like, the process is quickly recommended.
Furthermore, in the embodiment of the invention, the process recommendation is carried out by using a graph similarity measurement method, and the input data is graph structure data. The method for measuring the similarity of the graph can avoid the complicated steps of carrying out similarity measurement by classifying before violence searching in the prior art. In the embodiment, a target process recommendation model is directly used for process recommendation, so that violent search is avoided, and the process recommendation efficiency is improved.
As an optional application scenario of the process recommendation method in this embodiment, the process recommendation method is applied to an electronic device, such as a computer or a server. When parts need to be machined, inputting machining characteristics of a target product into electronic equipment, and determining a recommended process based on the machining characteristics and a target process recommendation model; and sending the recommended process to the part machining equipment so as to realize the machining of the parts. Or the input of the processing characteristics of the target product is input on the part processing equipment and is sent to the electronic equipment by the part processing equipment, and accordingly, the electronic equipment acquires the processing characteristics of the target product. When the electronic device is a server, the server may make process recommendations for a plurality of parts machining devices.
As another optional application scenario of the process recommendation method in this embodiment, the process recommendation method is applied to a part machining apparatus, for example, a numerical control machine. The method comprises the steps that a target process recommendation model is built in the part machining equipment, when a target product needs to be machined, a user inputs machining characteristics of the target product on the part machining equipment, and the part machining equipment determines a recommended process by executing the process recommendation method in the embodiment of the invention, so that the part machining equipment utilizes the recommended process to machine the target product.
It should be noted that the machining process determined by the target process recommendation model can be used as a reference, and a user can adjust the machining process based on the reference so as to better adapt to the machining of the product. For example, the machining process determined by the target process recommendation model is displayed on an interface of the electronic device, and the user adjusts the machining process in a manner of interacting with the electronic device, so that the machining process for machining the target product is determined. Of course, the target product can also be processed by directly using the processing technology determined by the target technology recommendation model.
The process recommendation method provided in the embodiment of the present invention may also be applied in other scenarios, which are only an example and do not limit the protection scope of the present invention, and the specific application scenario is set according to actual requirements.
In accordance with an embodiment of the present invention, there is provided a method for training a process recommendation model and a process recommendation method embodiment, it is noted that the steps illustrated in the flow chart of the accompanying drawings may be executed in a computer system such as a set of computer executable instructions, and that while a logical order is illustrated in the flow chart, in some cases, the steps illustrated or described may be executed in an order different than that shown herein.
In this embodiment, a method for training a process recommendation model is provided, which may be used in electronic devices, such as computers, numerically controlled machine tools, servers, and the like, fig. 1 is a flowchart of a method for training a process recommendation model according to an embodiment of the present invention, and as shown in fig. 1, the flowchart includes the following steps:
and S11, acquiring the sample processing characteristic sub-graph pair and the target similarity between the sample processing characteristic sub-graph pairs.
The target similarity is determined according to a similarity measurement mode corresponding to the attribute type of each node in the sample processing feature subgraph pair, and the attribute type comprises quantitative attributes and semantic attributes.
The sample processing feature sub-graph pair is a combination of two sample processing feature sub-graphs, for example, two sample processing feature sub-graphs corresponding to hole features constitute a sample processing feature sub-graph pair, and two sample processing feature sub-graphs corresponding to groove features constitute a sample processing feature sub-graph pair. The determination of the sample processing feature sub-graph is constructed by using process data of a processed product, for example, a CAM process file is obtained from a factory, and then the processing feature and the processing process data can be extracted by a process feature extraction algorithm, wherein the process feature extraction algorithm is specifically realized by a graph matching method. Of course, the process feature extraction algorithm is not limited to the above-described graph matching method, and may also be implemented in other ways.
Specifically, the machined features include hole features, slot features, and the like; therein, the pore characteristics may be subdivided into: fine size through holes, simple unthreaded through holes, fine size blind holes, simple sharp bottom blind holes, threaded holes and the like; the tank may be subdivided into: a through waist-shaped groove, a non-through waist-shaped groove, etc. The machining process steps include various steps, such as machining steps, tool types, tool selection rules, machining depths, machining methods, and the like.
It should be noted that the scope of the processing features and the processing techniques is not limited to the above description, and other processing features and processing techniques may also be set according to actual requirements. Depending, among other things, on the design requirements of the part processing equipment.
For example, fig. 2 shows an example of a sample processing feature sub-graph, which is an attribute feature belonging to a fine-scale via feature under a hole feature, the attribute feature including an aperture diameter, a magnification ratio, a precision, and a roughness. Wherein the sample processing feature sub-graph is determined based on relationships between processing features.
And the target similarity between the sample processing characteristic subgraph pairs is a label of a preset process recommendation model for subsequent training, and is determined by a similarity measurement mode corresponding to the attribute type of each node in the sample processing characteristic subgraph pairs. That is, different similarity measures are used for different attribute types. Specifically, the quantitative attributes, namely the nodes in the sample processing characteristic subgraph, are represented by numerical values; nodes in the semantic attribute and sample processing feature subgraph are represented by character strings, such as node Ra0.8 in FIG. 2.
For the similarity between the nodes with quantitative attributes, the difference value calculation can be directly carried out by utilizing the numerical values corresponding to the nodes. For the similarity between nodes of semantic attributes, the distance calculation between character strings corresponding to the nodes may be utilized, and so on. After obtaining the similarity between the nodes, the electronic device integrates all the similarities, for example, a weighted sum, an average value, and the like, to determine the target similarity.
Details about this step will be described later.
And S12, inputting the sample processing characteristic sub-graph pairs into a preset process recommendation model to obtain the prediction similarity.
The input of the preset process recommendation model is a pair of processing characteristic graphs, and the output is the prediction similarity between the pair of processing characteristic graphs. For the preset process recommendation model, vectorization is carried out on the relative sample processing characteristic subgraph pair, and then similarity calculation is carried out to obtain the prediction similarity. The similarity is calculated in the same manner as the target similarity in S11, that is, different similarity measures are used for different attribute types.
The specific structural details of the preset process recommendation model are not limited at all, and the preset process recommendation model is only required to be capable of determining the corresponding prediction similarity of the two input sample processing characteristic subgraphs.
Details about this step will be described later.
And S13, adjusting parameters of the preset process recommendation model based on the difference between the prediction similarity and the target similarity to determine the trained target process recommendation model.
And the electronic equipment performs loss function calculation by using the prediction similarity and the target similarity to determine loss, and then adjusts the parameters of the preset process recommendation model based on the determined loss. And determining the trained target process recommendation model by combining the loss function of the minimization model through multiple iterative cycles. The loss function is set according to actual requirements, and is not limited herein.
The training method for the process recommendation model provided by this embodiment implements instance description of process data by using a knowledge graph method, is used for actual retrieval based on a graph similarity measurement algorithm, and performs similarity calculation in combination with a similarity measurement mode corresponding to an attribute type of each node in a sample processing feature subgraph pair in an actual retrieval process, that is, by converting graph structure data of different node attribute types into a vector form and performing calculation between vectors, the calculation of the similarity between nodes of various types is ensured, the accuracy and recommendation efficiency of the trained target process recommendation model are improved, and accordingly, the efficiency of subsequently performing process recommendation by using the target process recommendation model is improved.
In this embodiment, a method for training a process recommendation model is provided, which can be used in electronic devices, such as computers, numerically controlled machine tools, servers, and the like, fig. 3 is a flowchart of a method for training a process recommendation model according to an embodiment of the present invention, and as shown in fig. 3, the flowchart includes the following steps:
and S21, acquiring the sample processing characteristic sub-graph pair and the target similarity between the sample processing characteristic sub-graph pairs.
The target similarity is determined according to a similarity measurement mode corresponding to the attribute type of each node in the sample processing feature subgraph pair, and the attribute type comprises quantitative attributes and semantic attributes.
Specifically, S21 includes:
and S211, obtaining a sample processing characteristic subgraph pair.
Please refer to the corresponding description of the embodiment S11 shown in fig. 1 for the obtaining method of the sample processing feature sub-graph pair, which is not described herein again.
And S212, determining corresponding attribute types based on the data types of the nodes in the sample processing characteristic subgraph pair.
After the electronic device obtains the sample processing characteristic sub-graph pair, by analyzing the data type of each node, for example, an integer type, a floating point type, a character type, and the like, the corresponding attribute type can be obtained after the data type is determined. If the attribute type is numerical, determining the attribute type as a quantitative attribute; and if the character type is the character type, determining that the attribute type is the semantic attribute.
And S213, determining the similarity between corresponding nodes in the sample processing characteristic subgraph pair by using the similarity measurement mode corresponding to the attribute type.
After the attribute type is determined, the electronic equipment calculates the similarity between the corresponding nodes by using a similarity measurement mode corresponding to the attribute type.
In some optional embodiments, the S213 includes:
(1) and when the attribute type is the quantitative attribute, performing similarity calculation on the numerical values of the corresponding nodes in the pair based on the sample processing characteristic subgraph to obtain the similarity.
(2) And when the attribute type is the semantic attribute, performing similarity calculation based on the editing distance between the character strings of the corresponding nodes in the sample processing characteristic subgraph pair to obtain the similarity.
For quantitative attributes, it can be described in terms of values, including aperture, magnification ratio, precision, roughness, groove length, groove width, and the like. Specifically, the similarity between nodes belonging to the quantitative attribute is calculated using the following formula:
Figure BDA0003634523370000101
wherein X, Y is two sample processing characteristic subgraphs,
Figure BDA0003634523370000102
respectively representing the i-th characteristic attribute of X, Y, max (C) i ) Processing the maximum value of the ith characteristic attribute, min (C) in the characteristic subgraphs for all samples i ) And processing the minimum value of the ith characteristic attribute in the characteristic subgraphs for all samples.
For semantic attributes, they are mainly described in terms of strings or character sets in the model design process, and these descriptions cannot be quantified. Such as: feature type, feature name, etc. The similarity of the two character strings of the 'simple sharp-bottom blind hole' and the 'fine-size sharp-bottom blind hole' is greater than that of the two character strings of the 'simple sharp-bottom blind hole' and the 'waist-shaped groove'. Because semantic attributes are mainly represented by character strings, the descriptions cannot be quantized, and the conventional semantic attribute measurement formula is directly judged by 0, 1, for example, the following formula is not accurate in semantic attribute judgment:
Figure BDA0003634523370000111
based on this, the edit distance is employed in the present embodiment to measure the similarity between two character strings. The edit distance is the shortest distance for changing one character string into another character string, and describes the closeness of the two character strings.
The calculation mode of the editing distance is set according to actual requirements, only the distance between two character strings needs to be calculated, and the similarity is represented by the distance between the character strings.
When the attribute type is a quantitative attribute, directly utilizing corresponding data to carry out similarity calculation so as to reduce data processing amount brought by data conversion; when the attribute type is the semantic attribute, the editing distance between the character strings is used for carrying out similarity calculation, and the editing distance is used for measuring the similarity between different words and describing the similarity between the two character strings, so that the similarity between nodes of the semantic attribute is described by using the editing distance, and the accuracy of the similarity calculation of the semantic attribute is improved.
And S214, fusing the similarity between corresponding nodes in the sample processing characteristic subgraph pair to determine the target similarity.
As described above, the fusion may use a weighted sum, an average value, and the like, that is, all the similarities calculated in S213 are fused to determine the target similarity.
In some alternative embodiments, the S214 includes:
(1) and acquiring the weight corresponding to each similarity.
(2) And performing weighted calculation based on the weight and the corresponding similarity to determine the target similarity.
The weight corresponding to each similarity is set according to actual requirements, and is not limited herein. Or, in order to reduce the complexity of data processing, different weights are set for different attribute types, that is, the weights of the same attribute type are the same. After the weight is determined, the electronic device may perform the calculation of the target similarity based on the following formula:
Figure BDA0003634523370000112
and all the similarities are fused in a weighting mode, and the target similarity can be determined by reducing the calculated amount, so that the data processing efficiency is improved.
And S22, inputting the sample processing characteristic sub-graph pairs into a preset process recommendation model to obtain the prediction similarity.
Please refer to S12 in fig. 1, which is not described herein again.
And S23, adjusting parameters of the preset process recommendation model based on the difference between the prediction similarity and the target similarity to determine the trained target process recommendation model.
Please refer to S13 in fig. 1, which is not described herein again.
According to the training method of the process recommendation model, the corresponding attribute types are obtained through the data types of the nodes in the sample processing characteristic subgraph pair, so that an accurate similarity measurement mode is determined, the similarities among the nodes are fused, and the accuracy of the determined target similarity is improved.
In this embodiment, a method for training a process recommendation model is provided, which can be used in electronic devices, such as computers, numerically controlled machine tools, servers, and the like, fig. 4 is a flowchart of a method for training a process recommendation model according to an embodiment of the present invention, and as shown in fig. 4, the flowchart includes the following steps:
and S31, acquiring the sample processing characteristic sub-graph pair and the target similarity between the sample processing characteristic sub-graph pairs.
The target similarity is determined according to a similarity measurement mode corresponding to the attribute type of each node in the sample processing feature subgraph pair, and the attribute type comprises quantitative attributes and semantic attributes.
Please refer to S21 in fig. 3 for details, which are not described herein.
And S32, inputting the sample processing characteristic sub-graph pairs into a preset process recommendation model to obtain the prediction similarity.
Specifically, S32 includes:
s321, vectorizing the sample processing feature sub-graph pair by using a vector module in the preset process recommendation model to obtain a sample processing feature vector pair with the same dimension.
The preset process recommendation model comprises a vector module and a similarity module, wherein the vector module is used for vectorizing the input sample processing feature subgraph to obtain a sample processing feature vector pair with the same dimension. That is, only on the basis of the same dimension, the subsequent similarity calculation can be performed. For example, if the vector dimensions of the hole feature and the groove feature are different, the similarity calculation cannot be performed using vectorization, resulting in a failure of the similarity calculation. Therefore, before the similarity calculation, the graph data is vectorized into vectors having the same dimension.
The working principle of the vector module is that each sample processing characteristic subgraph is vectorized, for example, each node in the subgraph is coded and then fused; or, coding each sample processing characteristic subgraph as a whole, and the like.
In some optional embodiments, the S321 includes:
(1) and vectorizing the nodes of the sample processing characteristic sub-graph pair by using a graph convolution network module in the vector module to obtain a node vector pair.
(2) And performing graph vectorization on the node vector pairs by using an attention network module in the vector module to obtain sample processing characteristic sub-graph vector pairs with the same dimension.
As shown in fig. 5, firstly, vectorizing, namely embedding nodes, of the sample processing feature subgraph pair by using a graph convolution network module to obtain a node vector pair; and then, carrying out graph-level vectorization, namely graph embedding on the node vector pairs by using an attention network module to obtain sample processing characteristic subgraph vector pairs. And subsequently, similarity calculation is carried out based on the sample processing feature vector pair to obtain the prediction similarity.
Vectorizing the nodes of the sample processing characteristic subgraph pair, namely embedding the nodes; and then, vectorization at the graph level, namely graph embedding is carried out to obtain a sample processing characteristic subgraph vector pair, and the obtained sample processing characteristic subgraph pair is sequentially vectorized according to the hierarchical relationship, so that the obtained sample processing characteristic relatively retains the information of each node and integrates the integral information of the sample processing characteristic subgraph, and the accuracy of the obtained sample processing characteristic vector pair is improved.
And S322, performing similarity calculation on the sample processing feature vector pair by using a similarity module in the preset process recommendation model and a similarity measurement mode corresponding to the attribute type of each node in the sample processing feature subgraph, and determining the prediction similarity.
The working principle of the similarity module is the same as the calculation method of the target similarity in S21 in the embodiment shown in fig. 3, and please refer to the corresponding description of S21 in the embodiment shown in fig. 3 for details, which is not repeated herein.
And S33, adjusting parameters of the preset process recommendation model based on the difference between the prediction similarity and the target similarity to determine the trained target process recommendation model.
Please refer to S13 in fig. 1, which is not described herein again.
In the training method of the process recommendation model provided by this embodiment, before the processing of the similarity module, the input sample processing feature sub-graph pair is processed into a sample processing feature vector pair with the same dimension, so as to ensure that the similarity can be calculated subsequently.
As a specific application example of the training method of the process recommendation model in this embodiment, as shown in fig. 6, the method includes:
(1) firstly, constructing a sample data set by using process data of a historical product to obtain a process knowledge graph;
(2) sampling the process knowledge graph to obtain two sample processing characteristic sub-graphs, and determining a sample processing characteristic sub-graph pair;
(3) calculating the target similarity aiming at each sample processing characteristic subgraph pair;
(4) and performing model training by using the sample processing characteristic subgraph and the corresponding target similarity to obtain a target process recommendation model. Wherein the training process mainly comprises the process shown in fig. 5.
After the target process recommendation model is obtained, when a subsequent machining process is determined, the machining feature subgraph can be obtained first, and then the target process recommendation model is used for calculating the similarity between the machining feature subgraph and each machining subgraph in the process knowledge graph, so that the first N machining feature subgraphs with the highest similarity are determined. And determining a target processing technology subgraph by utilizing the corresponding relation between the processing characteristic subgraphs and the processing technology subgraphs in the technology knowledge graph, thereby determining the recommended processing technology.
In this embodiment, a process recommendation method is provided, which may be used in electronic devices, such as computers, numerically controlled machine tools, servers, and the like, fig. 7 is a flowchart of a training method of a process recommendation model according to an embodiment of the present invention, and as shown in fig. 7, the flowchart includes the following steps:
and S41, acquiring the machining characteristics to be recommended.
The processing characteristics to be recommended may be input by the user on the electronic device, or may be obtained by the electronic device from other devices, and the source thereof is not limited. And the processing characteristics to be recommended are the processing characteristics of the target product.
For example, an input interface with processing features is displayed on an interface of an electronic device, and is divided according to levels, wherein the first level includes: hole features, slot features, etc.; the second level includes: subdivision under hole and slot features; the third level includes attribute features under the subdivision features.
And S42, constructing a processing feature sub-graph to be recommended based on the relation between the processing features.
After the electronic device obtains the processing features, the relationship between the processing features is determined accordingly. For example, the relationship between the processing features can be determined by inputting the processing features under the corresponding labels of the interface.
In some alternative embodiments, the S44 includes:
(1) and determining a father node corresponding to the machining features based on the relation among the machining features.
(2) And constructing a processing feature sub-graph to be recommended based on the father node and the processing feature.
And the parent node corresponding to the processing feature corresponds to the first level of the input interface, and the child nodes are sequentially confirmed downwards, so that a processing feature sub-graph to be recommended is constructed. And determining the father node by using the relation among the processing characteristics, and determining the child node on the basis, so that the accuracy of the processing characteristic subgraph to be recommended is improved.
And S43, inputting the processing feature sub-graph to be recommended and each processing feature sub-graph in the process knowledge graph into the target process recommendation model to obtain at least one target processing feature sub-graph with the highest similarity to the processing feature sub-graph to be recommended in the process knowledge graph.
The process knowledge graph comprises a processing characteristic subgraph and a processing process subgraph, the processing characteristic subgraph and the processing process subgraph have a corresponding relation, and the target process recommendation model is obtained by training according to the training method of the process recommendation model and is not repeated herein.
As shown in fig. 8, the subgraph framed by the rectangular frame represents a processing feature subgraph, and there is an association relationship between the processing feature subgraph and the processing technology subgraph, that is, there may be a many-to-many relationship, or a one-to-one relationship, or a many-to-one relationship between the processing feature subgraph and the processing technology subgraph in the technology knowledge graph, and so on. The corresponding relation between the processing characteristic subgraph and the processing technology subgraph is not limited at all, and the processing characteristic subgraph and the processing technology subgraph can be set according to actual requirements.
It should be noted that fig. 8 only shows the form of the process knowledge graph, and does not limit the content of each node; that is, the specific display contents of each node in fig. 8 are not limited at all.
The electronic device determines the first N target processing feature sub-graphs with the highest similarity by using the target process recommendation model, wherein the number of the target processing feature sub-graphs may be 1, 2, or multiple, and is not limited herein.
And S44, retrieving in the process knowledge map based on the target processing characteristic subgraph, and determining the target processing process subgraph corresponding to the target processing characteristic subgraph to determine the recommended processing process.
After obtaining the target processing characteristic subgraph, the electronic equipment searches in a process knowledge graph, for example, searches in a database of a knowledge graph data layer Neo4j, and searches in a Cypher search language. And searching each node matched with the processing characteristic subgraph to search out a target processing technology subgraph corresponding to the target processing characteristic subgraph.
The electronic equipment can display the searched target processing technology subgraph, namely, the target processing technology subgraph is displayed visually; and the user can adjust the target processing technology subgraph by combining with the actual requirement, so that the recommended processing technology is determined.
In the process recommendation method provided by the embodiment, the target process recommendation model obtained by training is used for determining at least one target processing characteristic subgraph, the target process recommendation model is processed based on graph data, and the similarity calculation is performed by distinguishing calculation based on the attribute types of the nodes, so that the determination efficiency of the target processing characteristic subgraph is improved, the process knowledge graph is retrieved on the basis, and the retrieval efficiency of the target processing characteristic subgraph is ensured. The knowledge graph-based process recommendation method has strong generalization, and the knowledge graph can be continuously modified and adjusted in the face of new industries so as to meet the requirements of different industries.
Referring to fig. 6, as a specific application example of the process recommendation method and the training method of the process recommendation model, for the machining recommendation model, the process recommendation model is divided into two stages, one is a training stage, and the other is a recommendation stage, where model inputs and outputs of the two stages are different, specifically:
(1) model training phase
The input of the training method is processing feature subgraph pairs, for example, a data set comprises processing feature data of 11 different holes and grooves, 2000 processing feature subgraphs are obtained, and every two subgraph pairs are obtained by sampling the processing feature subgraph pairs, wherein the total number of the processing feature subgraph pairs is 60000. Wherein, the training subgraph pair is divided, and the proportion of the training set and the test set is 70 percent and 30 percent.
Output of the training phase: after the model is trained by the training set data, a sub-graph pair similarity score is output on the test set, and an evaluation index is calculated according to the predicted similarity score and the actual similarity score, wherein the evaluation index is MSE (mean square error) and p @ k (the first k accuracy degrees). If the evaluation result meets the requirement, the model training is finished; if the evaluation result does not meet the requirement, the parameter of the model is required to be adjusted continuously until the evaluation requirement is met.
(2) Model recommendation phase
The input of the recommendation method is as follows: machining features such as pore size, aspect ratio, roughness, etc. The electronic equipment converts the input processing characteristics into processing characteristic subgraphs to be recommended according to the input processing characteristics, the processing characteristic subgraphs to be recommended are input into a target process recommendation model, subgraph pairs are formed by the processing characteristic subgraphs and all 2000 processing characteristic subgraphs, similarity degree score prediction is rapidly carried out on the model, and N processing characteristic subgraphs with the highest similarity degree are taken for recommendation.
And (3) outputting an algorithm: and N processing characteristic subgraphs with highest similarity.
The embodiment also provides a training device of a process recommendation model and a process recommendation device, which are used for implementing the above embodiments and preferred embodiments and are not described again after being described. As used below, the term "module" may be a combination of software and/or hardware that implements a predetermined function. Although the means described in the embodiments below are preferably implemented in software, an implementation in hardware, or a combination of software and hardware is also possible and contemplated.
The present embodiment provides a training apparatus for a process recommendation model, as shown in fig. 9, including:
a first obtaining module 51, configured to obtain a sample processing feature sub-graph pair and a target similarity between the sample processing feature sub-graph pairs, where the target similarity is determined according to a similarity measurement mode corresponding to an attribute type of each node in the sample processing feature sub-graph pair, and the attribute type includes a quantitative attribute and a semantic attribute;
the input module 52 is used for inputting the sample processing characteristic subgraph pair into a preset process recommendation model to obtain a prediction similarity;
and an adjusting module 53, configured to adjust parameters of the preset process recommendation model based on a difference between the predicted similarity and the target similarity, so as to determine a trained target process recommendation model.
The present embodiment provides a process recommendation apparatus, as shown in fig. 10, including:
the second obtaining module 61 is configured to obtain a machining feature to be recommended;
a building module 62, configured to build a to-be-recommended machining feature sub-graph based on the relationship between the machining features;
a recommending module 63, configured to input the processing feature sub-graph to be recommended and each processing feature sub-graph in a process knowledge graph into a target process recommendation model, to obtain at least one target processing feature sub-graph with the highest similarity to the processing feature sub-graph to be recommended in the process knowledge graph, where the process knowledge graph includes the processing feature sub-graph and the processing process sub-graph, the processing feature sub-graph and the processing process sub-graph have a corresponding relationship, and the target process recommendation model is obtained by training according to the training method of the process recommendation model in the first aspect of the present invention or any one of the embodiments of the first aspect;
and the retrieval module 64 is used for retrieving in the process knowledge graph based on the target processing characteristic subgraph, and determining a target processing process subgraph corresponding to the target processing characteristic subgraph so as to determine a recommended processing process.
The process recommendation model training device or process recommendation device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC circuit, a processor and a memory executing one or more software or fixed programs, and/or other devices capable of providing the above functions.
Further functional descriptions of the modules are the same as those of the corresponding embodiments, and are not repeated herein.
An embodiment of the present invention further provides an electronic device, which includes the training apparatus of the process recommendation model shown in fig. 9 or the process recommendation apparatus shown in fig. 10.
Referring to fig. 11, fig. 11 is a schematic structural diagram of an electronic device according to an alternative embodiment of the present invention, and as shown in fig. 11, the electronic device may include: at least one processor 71, such as a CPU (Central Processing Unit), at least one communication interface 73, memory 74, at least one communication bus 72. Wherein a communication bus 72 is used to enable the connection communication between these components. The communication interface 73 may include a Display (Display) and a Keyboard (Keyboard), and the optional communication interface 73 may also include a standard wired interface and a standard wireless interface. The Memory 74 may be a high-speed RAM Memory (volatile Random Access Memory) or a non-volatile Memory (non-volatile Memory), such as at least one disk Memory. The memory 74 may alternatively be at least one memory device located remotely from the processor 71. Wherein the processor 71 may be in connection with the apparatus described in fig. 9 or fig. 10, an application program is stored in the memory 74, and the processor 71 calls the program code stored in the memory 74 for performing any of the above-mentioned method steps.
The communication bus 72 may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The communication bus 72 may be divided into an address bus, a data bus, a control bus, and the like. For ease of illustration, only one thick line is shown in FIG. 11, but this is not intended to represent only one bus or type of bus.
The memory 74 may include a volatile memory (RAM), such as a random-access memory (RAM); the memory may also include a non-volatile memory (english: non-volatile memory), such as a flash memory (english: flash memory), a hard disk (english: hard disk drive, abbreviated: HDD) or a solid-state drive (english: SSD); the memory 74 may also comprise a combination of memories of the kind described above.
The processor 71 may be a Central Processing Unit (CPU), a Network Processor (NP), or a combination of CPU and NP.
The processor 71 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a Programmable Logic Device (PLD), or a combination thereof. The PLD may be a Complex Programmable Logic Device (CPLD), a field-programmable gate array (FPGA), a General Array Logic (GAL), or any combination thereof.
Optionally, the memory 74 is also used for storing program instructions. The processor 71 may call program instructions to implement a training method of a process recommendation model as shown in any of the embodiments of the present application, or a process recommendation method as shown in any of the embodiments.
The embodiment of the invention also provides a non-transitory computer storage medium, wherein the computer storage medium stores computer executable instructions, and the computer executable instructions can execute the training method of the process recommendation model or the process recommendation method in any method embodiment. The storage medium may be a magnetic Disk, an optical Disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a Flash Memory (Flash Memory), a Hard Disk Drive (Hard Disk Drive, abbreviated as HDD), or a Solid State Drive (SSD); the storage medium may also comprise a combination of memories of the kind described above.
Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations fall within the scope defined by the appended claims.

Claims (10)

1. A training method of a process recommendation model is characterized by comprising the following steps:
acquiring a sample processing characteristic subgraph pair and target similarity between the sample processing characteristic subgraph pair, wherein the target similarity is determined according to a similarity measurement mode corresponding to an attribute type of each node in the sample processing characteristic subgraph pair, and the attribute type comprises a quantitative attribute and a semantic attribute;
inputting the sample processing characteristic subgraph pair into a preset process recommendation model to obtain prediction similarity;
and adjusting parameters of the preset process recommendation model based on the difference between the predicted similarity and the target similarity so as to determine a trained target process recommendation model.
2. The training method of claim 1, wherein obtaining the target similarity between the sample processing feature sub-graph pairs comprises:
determining the corresponding attribute type based on the data type of each node in the sample processing characteristic subgraph pair;
determining the similarity between corresponding nodes in the sample processing characteristic subgraph pair by using a similarity measurement mode corresponding to the attribute type;
and fusing the similarity between corresponding nodes in the sample processing characteristic subgraph pair to determine the target similarity.
3. The training method according to claim 2, wherein the determining the similarity between corresponding nodes in the sample processing feature subgraph pair by using the similarity measure corresponding to the attribute type comprises:
when the attribute type is the quantitative attribute, performing similarity calculation based on the numerical values of the corresponding nodes in the sample processing characteristic subgraph pair to obtain the similarity;
and when the attribute type is the semantic attribute, performing similarity calculation based on the editing distance between the character strings of the corresponding nodes in the sample processing characteristic subgraph pair to obtain the similarity.
4. The training method of claim 2, wherein the fusing the similarity between corresponding nodes in the sample processing feature subgraph pair to determine the target similarity comprises:
acquiring weights corresponding to the similarity;
and performing weighted calculation based on the weight and the corresponding similarity to determine the target similarity.
5. The training method according to any one of claims 1 to 4, wherein the step of inputting the sample processing feature subgraph pair into a preset process recommendation model to obtain a prediction similarity comprises the steps of:
vectorizing the sample processing characteristic sub-graph pair by using a vector module in the preset process recommendation model to obtain a sample processing characteristic vector pair with the same dimension;
and performing similarity calculation on the sample processing feature vector pair by utilizing a similarity module in the preset process recommendation model and a similarity measurement mode corresponding to the attribute type of each node in the sample processing feature subgraph to determine the prediction similarity.
6. The training method of claim 5, wherein the vectorizing the sample processing feature sub-graph pair by using a vector module in the preset process recommendation model to obtain a sample processing feature vector pair with the same dimension comprises:
vectorizing the nodes of the sample processing characteristic subgraph pair by using a graph convolution network module in the vector module to obtain a node vector pair;
and carrying out graph vectorization on the node vector pair by using an attention network module in the vector module to obtain the sample processing characteristic sub-graph vector pair with the same dimension.
7. A method of process recommendation, comprising:
acquiring a processing characteristic to be recommended;
constructing a processing feature sub-graph to be recommended based on the relation between the processing features;
inputting each processing feature sub-graph in the processing feature sub-graph to be recommended and the process knowledge graph into a target process recommendation model to obtain at least one target processing feature sub-graph with the highest similarity to the processing feature sub-graph to be recommended in the process knowledge graph, wherein the process knowledge graph comprises the processing feature sub-graph and the processing process sub-graph, the processing feature sub-graph and the processing process sub-graph have a corresponding relation, and the target process recommendation model is obtained by training according to the training method of the process recommendation model of any one of claims 1-6;
and searching in the process knowledge graph based on the target processing characteristic subgraph, and determining a target processing process subgraph corresponding to the target processing characteristic subgraph to determine a recommended processing process.
8. The process recommendation method according to claim 7, wherein the building of the processing feature subgraph to be recommended based on the relation between the processing features comprises:
determining a father node corresponding to the machining features based on the relation among the machining features;
and constructing the processing feature subgraph to be recommended based on the father node and the processing feature.
9. An electronic device, comprising:
a memory and a processor, the memory and the processor being communicatively connected to each other, the memory having stored therein computer instructions, the processor executing the computer instructions to perform the method of training the process recommendation model of any one of claims 1-6 or to perform the method of process recommendation of claim 7 or 8.
10. A computer-readable storage medium storing computer instructions for causing a computer to perform the method of training a process recommendation model according to any one of claims 1-6, or to perform the method of process recommendation according to claim 7 or 8.
CN202210501295.5A 2022-05-09 2022-05-09 Training method of process recommendation model, process recommendation method and electronic equipment Pending CN114880457A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210501295.5A CN114880457A (en) 2022-05-09 2022-05-09 Training method of process recommendation model, process recommendation method and electronic equipment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210501295.5A CN114880457A (en) 2022-05-09 2022-05-09 Training method of process recommendation model, process recommendation method and electronic equipment

Publications (1)

Publication Number Publication Date
CN114880457A true CN114880457A (en) 2022-08-09

Family

ID=82674649

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210501295.5A Pending CN114880457A (en) 2022-05-09 2022-05-09 Training method of process recommendation model, process recommendation method and electronic equipment

Country Status (1)

Country Link
CN (1) CN114880457A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116628228A (en) * 2023-07-19 2023-08-22 安徽思高智能科技有限公司 RPA flow recommendation method and computer readable storage medium

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116628228A (en) * 2023-07-19 2023-08-22 安徽思高智能科技有限公司 RPA flow recommendation method and computer readable storage medium
CN116628228B (en) * 2023-07-19 2023-09-19 安徽思高智能科技有限公司 RPA flow recommendation method and computer readable storage medium

Similar Documents

Publication Publication Date Title
US20230100376A1 (en) Text sentence processing method and apparatus, computer device, and storage medium
US11741361B2 (en) Machine learning-based network model building method and apparatus
US20230039496A1 (en) Question-and-answer processing method, electronic device and computer readable medium
CN111143569A (en) Data processing method and device and computer readable storage medium
CN112306494A (en) Code classification and clustering method based on convolution and cyclic neural network
CN113791757A (en) Software requirement and code mapping method and system
CN114880457A (en) Training method of process recommendation model, process recommendation method and electronic equipment
CN114330717A (en) Data processing method and device
CN113869609A (en) Method and system for predicting confidence of frequent subgraph of root cause analysis
CN113642727A (en) Training method of neural network model and processing method and device of multimedia information
CN113254649A (en) Sensitive content recognition model training method, text recognition method and related device
CN113821657A (en) Artificial intelligence-based image processing model training method and image processing method
JP2022166126A (en) Training method for document processing model, device, apparatus, storage medium, and program
Singaravelan et al. Analysis of classification algorithms on different datasets
CN114254622A (en) Intention identification method and device
CN113065321A (en) User behavior prediction method and system based on LSTM model and hypergraph
CN113590774A (en) Event query method, device and storage medium
CN111291182A (en) Hotspot event discovery method, device, equipment and storage medium
Liu et al. Table-Top Scene Analysis Using Knowledge-Supervised MCMC
CN113284256B (en) MR (magnetic resonance) mixed reality three-dimensional scene material library generation method and system
JP7444625B2 (en) question answering device
CN112364666B (en) Text characterization method and device and computer equipment
KR101983493B1 (en) System and method for visualizing and interpreting property included learned entity representation from natural language text
CN115544522B (en) Source code vulnerability detection method and device, electronic equipment and storage medium
CN114385831B (en) Knowledge-graph relation prediction method based on feature extraction

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination