Industrial internet system based on data mining technology

Abstract

The invention relates to the technical field of data mining, in particular to an industrial internet system based on a data mining technology, which comprises a data acquisition module, a node construction module, a node quality analysis module and a model training module, and is characterized in that the data acquisition module acquires equipment data and sensor data in the industry; the node construction module is used for constructing nodes according to the acquired data and performing data enhancement on the nodes to form a positive sample pair and a negative sample pair; the node quality analysis module is used for carrying out feature difference calculation on the positive sample pair, repairing the node according to a feature difference result, and generating a complete positive sample pair; feature difference calculation is carried out on the negative samples, nodes are repaired according to the result of the feature difference, and a complete negative sample pair is generated; the model training module initializes parameters in the graph neural network model and calculates the similarity of the complete positive sample pair and the similarity of the complete negative sample pair so as to optimize the parameters in the graph neural network model.

Description

Technical Field

[0001] The present invention relates to the field of data mining technology, and specifically to an industrial Internet system based on data mining technology. Background Art

[0002] Traditional industrial Internet systems often face multiple technical problems when processing device data and sensor data, which directly affect the training effect and actual application performance of the model.

[0003] First, the problem of data quality is particularly prominent. In traditional systems, equipment data and sensor data are usually used directly for model training without careful quality assessment and repair. Since the data collection process in industrial environments may have abnormal conditions such as noise and missing values, the node feature information in the positive sample pairs and negative sample pairs is not consistent or the difference is not obvious. This data quality problem will cause the model to be unable to accurately capture the internal true behavior patterns of the equipment or sensor during the training process, thereby affecting the accuracy and robustness of the model. For example, in anomaly detection, low-quality positive sample pairs may cause the model to falsely report normal behavior as abnormal, increasing unnecessary maintenance costs and production interruptions; in fault prediction, the model may delay fault detection due to the inability to accurately identify the equipment status, missing the best preventive maintenance opportunity.

[0004] Secondly, traditional systems lack effective data enhancement methods in the generation and processing of sample pairs. Due to the diversity and complexity of industrial data, the lack of diverse sample pairs will make the model's generalization ability weak when facing new data, and it will be difficult to cope with various situations in actual production. For example, in state recognition, the model may not be able to accurately judge the actual working state of the equipment due to the lack of diversity in the training data, resulting in poor production monitoring results; in equipment optimization scheduling, the model may produce misscheduling due to the inability to effectively distinguish the states of different equipment, reducing production efficiency.

[0005] In addition, in the design of loss functions and parameter optimization, traditional systems often fail to fully consider the contrast balance between positive and negative sample pairs, which can easily lead to inaccurate distinction between positive and negative samples during the training process, affecting the final performance of the model. For example, in state recognition, the imbalance of similarity can cause the model to deviate when judging the state of the equipment; in fault prediction, the inaccuracy of similarity can make the model unable to effectively distinguish between normal and faulty states, thus affecting the reliability of the prediction.

[0006] In summary, the technical problems existing in traditional industrial Internet systems in terms of data quality, sample pair generation and loss function optimization limit the training effect and actual application performance of the model, and urgently need to be improved through more meticulous and scientific methods. Summary of the invention

[0007] The purpose of the present invention is to provide an industrial Internet system based on data mining technology to solve the problems raised in the above background technology.

[0008] To achieve the above object, the present invention provides the following technical solution: an industrial Internet system based on data mining technology, comprising a data acquisition module, a node construction module, a node quality analysis module and a model training module, wherein: The data acquisition module acquires equipment data and sensor data in the industry; The node construction module constructs nodes according to the device data and the sensor data, performs different data enhancements on the same node to form a positive sample pair; and performs the same data enhancement on different nodes to form a negative sample pair; The node quality analysis module obtains the features of the nodes in the positive sample pair and the negative sample pair, calculates the feature difference of the features of the same node in the positive sample pair, determines the node quality of the same node in the positive sample pair according to the result of the feature difference, and repairs the nodes of the positive sample pair according to the node quality to generate a complete positive sample pair; calculates the feature difference of the features of different nodes in the negative sample pair, determines the node quality of different nodes in the negative sample pair according to the result of the feature difference, and repairs the nodes of the negative sample pair according to the node quality to generate a complete negative sample pair; The model training module initializes the parameters in the graph neural network model, calculates the similarity of complete positive sample pairs, and calculates the similarity of complete negative samples; through back propagation and gradient descent, the similarity between complete positive sample pairs is maximized, and the similarity between complete negative sample pairs is minimized to optimize the parameters in the graph neural network model.

[0009] As a further improvement of the technical solution, the process of forming positive sample pairs and negative sample pairs by the node construction module specifically includes: The equipment and sensors in the industry are regarded as nodes in the graph, and each node is represented by a feature vector, which contains all the features extracted from the equipment data and sensor data; Perform different data enhancements on the same node in the graph to generate different versions of the node; perform different data enhancements on the same node in the graph through feature masking and feature perturbation data enhancement methods; for feature masking, determine the feature ratio that needs to be masked, randomly select features of the feature ratio, and set the value of the feature to a random value to simulate data noise; for feature perturbation, determine the distribution parameters of the data noise, and add the random noise of the distribution to each feature value to simulate the error in the data acquisition process; form positive sample pairs from different data enhancement versions of the same node; Two different nodes are selected from the graph, and the two nodes are not directly connected neighbor nodes in the original graph; the same data enhancement method is used for the two different nodes to generate data enhanced versions of the two different nodes, wherein the data enhancement method includes a feature mask; and the data enhanced versions of the two different nodes are used as negative sample pairs.

[0010] As a further improvement of the technical solution, the node quality analysis module includes a positive sample pair analysis unit, and the process of generating a complete positive sample pair by the positive sample pair analysis unit specifically includes: Compare the feature vectors of two different versions of the same node in the positive sample pair and use different distance metrics to calculate the feature difference between the feature vectors; According to the result of feature difference, the quality of the same node in the positive sample pair is evaluated; the smaller the feature difference, the higher the node quality; by setting a positive sample pair node threshold, it is determined whether the node in the positive sample pair is a low-quality node, and the node whose feature difference is greater than the positive sample pair node threshold is regarded as a low-quality node; The low-quality nodes in the positive sample pair are repaired to generate a complete positive sample pair. By re-enhancing the nodes, different methods are selected until the feature difference meets the threshold requirements of the positive sample pair nodes; the low-quality nodes in the repaired positive sample pair are filled into the positive sample pair to generate a complete positive sample pair.

[0011] As a further improvement of the technical solution, the node quality analysis module includes a negative sample pair analysis unit, and the process of generating a complete negative sample pair by the negative sample pair analysis unit specifically includes: Compare the feature vectors of the two enhanced versions of different nodes in the negative sample pair, and use the distance metric method to calculate the feature difference between the feature vectors; According to the result of feature difference, the quality of different nodes in the negative sample pair is evaluated. The greater the feature difference, the higher the node quality. By setting a negative sample pair node threshold, it is determined whether the node in the negative sample pair is a low-quality node. The node whose feature difference is less than the negative sample pair node threshold is regarded as a low-quality node. The low-quality nodes in the negative sample pair are repaired to generate a complete negative sample pair. The feature value is adjusted by adding noise to increase the feature difference, so as to perform the repair. The low-quality nodes in the repaired negative sample pair are filled into the negative sample pair to generate a complete negative sample pair.

[0012] As a further improvement of the technical solution, the process of optimizing the parameters of the graph neural network model by the model training module specifically includes: Select a graph neural network model and initialize the parameters in the graph neural network model, where the initialization method includes random initialization; Calculate the similarity of the complete positive sample pair and the complete negative sample pair according to the distance measurement method used in the node quality analysis module; use the calculated distance as the similarity of the corresponding sample pair; Define the loss function, and calculate the loss value under the parameters of the graph neural network model according to the loss function, by dividing the similarity of the complete positive sample pair by the adjustment parameter and taking the exponential value; dividing the similarity of the complete negative sample pair by the adjustment parameter and taking the exponential value; and adding the above exponential values ​​to obtain the sum of the exponential values; adding the exponential value of the complete positive sample similarity to the sum of the exponential values ​​of the complete negative sample similarity to obtain the denominator, and calculating the proportion of the exponential value of the complete positive sample similarity in the sum of the exponential values, and taking the negative logarithm to obtain the final loss value; the adjustment parameter is to adjust the distribution of the similarity to balance the contrast between positive and negative samples; Calculate the gradient of the loss value with respect to the parameters of the graph neural network model, and use the gradient descent algorithm to update the model parameters so that the similarity between complete positive sample pairs is maximized and the similarity between complete negative sample pairs is minimized to optimize the parameters in the graph neural network model.

[0013] Compared with the prior art, the present invention has the following beneficial effects: This industrial Internet system based on data mining technology significantly improves the data quality used to train graph neural network models by performing detailed quality assessment and repair of positive and negative sample pairs. Specifically, for low-quality nodes in the positive sample pairs, the system compares the feature differences of the same node under different data enhancement versions. If the feature difference exceeds the preset positive sample pair node threshold, the node will be marked as a low-quality node and will be re-enhanced until the quality requirements are met. This process ensures the consistency and reliability of the nodes in the positive sample pairs, so that the model can more accurately capture the internal true behavior patterns of the equipment or sensors during training. For example, in anomaly detection, high-quality positive sample pairs enable the model to more accurately identify the device behavior under normal conditions, significantly reducing the false alarm rate. In fault prediction, potential faults can be discovered earlier, so that preventive measures can be taken in advance. Overall, the repair process of the node quality analysis module effectively improves the quality of the data set and provides a reliable data foundation for model training.

[0014] In addition, by initializing the parameters in the graph neural network model and calculating the similarity of the complete positive sample pairs and negative sample pairs, and then optimizing the model parameters through back propagation and gradient descent algorithms, the similarity between positive sample pairs is maximized and the similarity between negative sample pairs is minimized. This process significantly enhances the model's discrimination ability and learning efficiency. In practical applications, the maximization of the similarity of positive sample pairs ensures that the model can better understand the consistency of the performance of the same device or sensor under different conditions, thereby improving the accuracy of the model. For example, in state recognition, the model can more accurately determine whether the equipment is in normal working condition, thereby improving the monitoring effect of the production process. Minimizing the similarity of negative sample pairs helps the model distinguish different devices or sensors, enhancing the robustness and generalization ability of the system. For example, in equipment optimization scheduling, the model can allocate resources and tasks more accurately, avoid misscheduling caused by similar equipment states, and significantly improve production efficiency. In general, the optimization process of the model training module not only improves the real-time responsiveness and accuracy of the industrial Internet system, but also enhances the system's prediction and decision support capabilities, which has an important impact on the optimization of the entire industrial production process. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic diagram of the overall module of the present invention; Figure 2 It is a schematic diagram of a node quality analysis module unit of the present invention.

[0016] In the figure: 100, data acquisition module; 200, node construction module; 300, node quality analysis module; 301, positive sample pair analysis unit; 302, negative sample pair analysis unit; 400, model training module. DETAILED DESCRIPTION

[0017] The following will be combined with the accompanying drawings in the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] Next, see Figure 1-Figure 2 The present invention provides a technical solution: an industrial Internet system based on data mining technology, including a data acquisition module 100, a node construction module 200, a node quality analysis module 300 and a model training module 400.

[0019] The data acquisition module 100 acquires corresponding device data and sensor data from the equipment and sensors in the industry. The device data includes operating status data based on time series (such as pressure and vibration at each time point) and sensor data includes environmental information (such as temperature and humidity). The device operating status information and environmental information in the device data and sensor data are standardized respectively to ensure that the data has a unified scale, thereby providing a guarantee for building reliable node and sample pairs.

[0020] The node construction module 200 constructs nodes according to the device data and the sensor data, performs different data enhancements on the same node to form a positive sample pair, and performs the same data enhancements on different nodes to form a negative sample pair, wherein the process of forming the positive sample pair and the negative sample pair specifically includes: Each device and sensor in the industry is considered as a node in the graph, which has device data and sensor data associated with it; features are extracted from the device data and sensor data. For time series data, statistical features (such as mean, variance, maximum value, minimum value) and time domain and frequency domain features (such as Fourier transform, wavelet transform) are used; each node is represented by a feature vector, which contains all features extracted from the device data and sensor data; Different data enhancements are performed on the same node in the graph to generate different versions of the node. The purpose of data enhancement is to increase the diversity of data so that the model can generalize better. Different data enhancements are performed on the same node in the graph through feature masking and feature perturbation data enhancement methods. For feature masking, first determine the feature ratio that needs to be masked, such as 20%, and then randomly select features of this ratio and set the value of the feature to 0 or a random value to simulate data missing or noise. For feature perturbation, first determine the distribution parameters of the noise, such as the mean of Gaussian noise is 0 and the standard deviation is 0.1, and then add the random noise of this distribution to each feature value to simulate the error in the data acquisition process. Positive pairs are formed by different data-augmented versions of the same node, where the data-augmented versions, although different in features, still essentially represent the same physical device or sensor; Select two different nodes from the graph, ensuring that the two nodes are not directly connected neighbor nodes in the original graph, or they have significant differences in the feature space; the nodes are selected to ensure that there is enough difference between them to form a valid negative sample pair in the subsequent steps; The same data augmentation method is used for the two different nodes to generate data augmented versions of the two different nodes, where the data augmentation method includes a feature mask; although the two nodes have undergone the same data augmentation method, there are still significant differences between the generated augmented versions due to their different original features or topological structures; therefore, the data augmented versions of these different nodes are regarded as negative sample pairs; negative sample pairs are used in contrastive learning to emphasize the differences between different nodes, helping the model to better learn node representation and generalization capabilities.

[0021] The positive sample pair analysis unit 301 in the node quality analysis module 300 calculates the feature difference of the features of the same node in the positive sample pair, determines the node quality of the same node in the positive sample pair according to the result of the feature difference, and repairs the positive sample pair node according to the node quality to generate a complete positive sample pair, which specifically includes: Compare the feature vectors of two different versions of the same node in the positive sample pair, and use different distance metrics, such as Euclidean distance, cosine similarity, etc., to calculate the feature difference between the feature vectors; According to the result of feature difference, the quality of the same node in the positive sample pair is evaluated; the smaller the feature difference, the higher the node quality, because it means that the data-enhanced version can still retain the main feature information of the original node; by setting a positive sample pair node threshold, it is determined whether the node in the positive sample pair is a low-quality node. For example, a node with a feature difference greater than the positive sample pair node threshold of 0.2 is considered a low-quality node; Repair the low-quality nodes in the positive sample pair to generate a complete positive sample pair. By re-enhancing the nodes, different methods are selected until the feature difference meets the threshold requirements of the positive sample pair nodes. If the feature difference is caused by the missing or abnormality of certain feature values, the node quality is improved by supplementing the missing data or correcting the abnormal data. Fill the low-quality nodes in the repaired positive sample pair into the positive sample pair to generate a complete positive sample pair.

[0022] The positive sample pair analysis unit 301 ensures that the node feature differences in the positive sample pairs are small and retains the main feature information of the nodes by evaluating and repairing the nodes in the positive sample pairs; this helps to more accurately learn the similarities of the positive sample pairs in the subsequent model training process, improve the training effect and reliability of the model, solve the problem of excessive differences in node features in the positive sample pairs, ensure the quality of the positive sample pairs, and thus help the model better learn the feature representation of the positive sample pairs.

[0023] The negative sample pair analysis unit 302 in the node quality analysis module 300 calculates feature differences of features of different nodes in the negative sample pair, determines the node quality of different nodes in the negative sample pair according to the result of the feature difference, and repairs the negative sample pair nodes according to the node quality to generate a complete negative sample pair, which specifically includes: Compare the feature vectors of the two enhanced versions of different nodes in the negative sample pair, and use a distance metric method (such as Euclidean distance, cosine similarity) to calculate the feature difference between the feature vectors; According to the result of feature difference, the quality of different nodes in the negative sample pair is evaluated. The larger the feature difference, the higher the node quality, because it means that the feature information of two different nodes is significantly different. By setting a negative sample pair node threshold, it is determined whether the node in the negative sample pair is a low-quality node. For example, a node with a feature difference less than 0.8 is considered a low-quality node. The low-quality nodes in the negative sample pair are repaired to generate a complete negative sample pair, and the feature difference is increased by adjusting the feature value; for example, the adjustment can be made by adding noise or selecting more representative features; the low-quality nodes in the repaired negative sample pair are filled into the negative sample pair to generate a complete negative sample pair.

[0024] The negative sample pair analysis unit 302 ensures that the node feature differences in the negative sample pairs are significant and enhances the contrast between different nodes by evaluating and repairing the nodes in the negative sample pairs. This helps the model to learn the differences in the negative sample pairs more accurately during the training process, improves the training effect and generalization ability of the model, solves the problem that the node feature differences in the negative sample pairs are not significant, ensures the quality of the negative sample pairs, and thus helps the model to better learn the differences in the negative sample pairs.

[0025] The model training module 400 initializes the parameters in the graph neural network model, and calculates the similarity of the complete positive sample pairs, and calculates the similarity of the complete negative samples; through back propagation and gradient descent, the similarity between the complete positive sample pairs is maximized, and the similarity between the complete negative sample pairs is minimized to optimize the parameters in the graph neural network model, specifically including: Select a graph neural network model, such as a convolutional network and an attention network; initialize the parameters in the graph neural network model, including the weight matrix and bias term, where the initialization method includes random initialization; Calculate the similarity of the complete positive sample pair and the similarity of the complete negative sample according to the distance measurement method used in the node quality analysis module 300; use the calculated distance as the similarity of the corresponding sample pair; Define the loss function, and calculate the loss value under the parameters of the graph neural network model according to the loss function, by dividing the similarity of the complete positive sample pair by the adjustment parameter and taking the exponential value; dividing the similarity of the complete negative sample pair by the adjustment parameter and taking the exponential value; and adding the above exponential values ​​to get the sum of the exponential values; adding the exponential value of the complete positive sample similarity to the sum of the exponential values ​​of the complete negative sample similarity to get the denominator, and calculating the proportion of the exponential value of the complete positive sample similarity in the sum of the exponential values, and taking the negative logarithm to get the final loss value; the adjustment parameter is to adjust the distribution of the similarity, so as to better balance the contrast between positive and negative samples; Calculate the gradient of the loss value with respect to the parameters of the graph neural network model, and use a gradient descent algorithm (such as SGD, Adam, etc.) to update the model parameters so that the similarity between complete positive sample pairs is maximized and the similarity between complete negative sample pairs is minimized to optimize the parameters in the graph neural network model; The model is optimized by maximizing the similarity between complete positive sample pairs and minimizing the similarity between complete negative sample pairs. The similarity between a node and multiple negative sample nodes is calculated and compared with the similarity of the positive sample nodes to ensure that the feature representations of complete positive sample pairs are closer, while the feature representations of complete negative sample pairs are farther apart. This allows the graph neural network model to better distinguish the feature representations of different nodes during training. The optimized model has stronger generalization and robustness and can perform well on new data. It solves the problem that the model cannot effectively distinguish between positive and negative sample pairs during training. By optimizing the loss function, the model can more accurately learn the feature representations of the nodes, thereby improving the performance and generalization ability of the model.

[0026] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. The industrial Internet system based on data mining technology is characterized by: The system comprises a data acquisition module (100), a node construction module (200), a node quality analysis module (300) and a model training module (400), wherein: The data acquisition module (100) acquires equipment data and sensor data in the industry; The node construction module (200) constructs nodes according to the device data and the sensor data, performs different data enhancements on the same node therein to form a positive sample pair; and performs the same data enhancements on different nodes therein to form a negative sample pair; The node quality analysis module (300) obtains the features of the nodes in the positive sample pair and the negative sample pair, calculates the feature difference of the features of the same node in the positive sample pair, determines the node quality of the same node in the positive sample pair according to the result of the feature difference, and repairs the nodes of the positive sample pair according to the node quality to generate a complete positive sample pair; calculates the feature difference of the features of different nodes in the negative sample pair, determines the node quality of different nodes in the negative sample pair according to the result of the feature difference, and repairs the nodes of the negative sample pair according to the node quality to generate a complete negative sample pair; The model training module (400) initializes the parameters in the graph neural network model, and calculates the similarity of complete positive sample pairs, as well as the similarity of complete negative samples; through back propagation and gradient descent, the similarity between complete positive sample pairs is maximized, and the similarity between complete negative sample pairs is minimized, so as to optimize the parameters in the graph neural network model.

2. The industrial Internet system based on data mining technology according to claim 1 is characterized in that: The process of forming positive sample pairs and negative sample pairs by the node construction module (200) specifically includes: The equipment and sensors in the industry are regarded as nodes in the graph, and each node is represented by a feature vector, which contains all the features extracted from the equipment data and sensor data; Perform different data enhancements on the same node in the graph to generate different versions of the node; perform different data enhancements on the same node in the graph through feature masking and feature perturbation data enhancement methods; for feature masking, determine the feature ratio that needs to be masked, randomly select features of the feature ratio, and set the value of the feature to a random value to simulate data noise; for feature perturbation, determine the distribution parameters of the data noise, and add the random noise of the distribution to each feature value to simulate the error in the data acquisition process; form positive sample pairs from different data enhancement versions of the same node; Two different nodes are selected from the graph, and the two nodes are not directly connected neighbor nodes in the original graph; the same data enhancement method is used for the two different nodes to generate data enhanced versions of the two different nodes, wherein the data enhancement method includes a feature mask; and the data enhanced versions of the two different nodes are used as negative sample pairs.

3. The industrial Internet system based on data mining technology according to claim 2 is characterized in that: The node quality analysis module (300) comprises a positive sample pair analysis unit (301), and the process of generating a complete positive sample pair by the positive sample pair analysis unit (301) specifically comprises: Compare the feature vectors of two different versions of the same node in the positive sample pair and use different distance metrics to calculate the feature difference between the feature vectors; According to the result of feature difference, the quality of the same node in the positive sample pair is evaluated; the smaller the feature difference, the higher the node quality; by setting a positive sample pair node threshold, it is determined whether the node in the positive sample pair is a low-quality node, and the node whose feature difference is greater than the positive sample pair node threshold is regarded as a low-quality node; The low-quality nodes in the positive sample pair are repaired to generate a complete positive sample pair. By re-enhancing the nodes, different methods are selected until the feature difference meets the threshold requirements of the positive sample pair nodes; the low-quality nodes in the repaired positive sample pair are filled into the positive sample pair to generate a complete positive sample pair.

4. The industrial Internet system based on data mining technology according to claim 3 is characterized in that: The node quality analysis module (300) comprises a negative sample pair analysis unit (302), and the process of generating a complete negative sample pair by the negative sample pair analysis unit (302) specifically comprises: Compare the feature vectors of the two enhanced versions of different nodes in the negative sample pair, and use the distance metric method to calculate the feature difference between the feature vectors; According to the result of feature difference, the quality of different nodes in the negative sample pair is evaluated. The greater the feature difference, the higher the node quality. By setting a negative sample pair node threshold, it is determined whether the node in the negative sample pair is a low-quality node. The node whose feature difference is less than the negative sample pair node threshold is regarded as a low-quality node. The low-quality nodes in the negative sample pair are repaired to generate a complete negative sample pair. The feature value is adjusted by adding noise to increase the feature difference, so as to perform the repair. The low-quality nodes in the repaired negative sample pair are filled into the negative sample pair to generate a complete negative sample pair.

5. The industrial Internet system based on data mining technology according to claim 4 is characterized in that: The process of optimizing the graph neural network model parameters by the model training module (400) specifically includes: Select a graph neural network model and initialize the parameters in the graph neural network model, where the initialization method includes random initialization; Calculating the similarity of the complete positive sample pair and the similarity of the complete negative sample according to the distance measurement method used in the node quality analysis module (300); using the calculated distance as the similarity of the corresponding sample pair; Define the loss function, and calculate the loss value under the parameters of the graph neural network model according to the loss function, by dividing the similarity of the complete positive sample pair by the adjustment parameter and taking the exponential value; dividing the similarity of the complete negative sample pair by the adjustment parameter and taking the exponential value; and adding the above exponential values ​​to obtain the sum of the exponential values; adding the exponential value of the complete positive sample similarity to the sum of the exponential values ​​of the complete negative sample similarity to obtain the denominator, and calculating the proportion of the exponential value of the complete positive sample similarity in the sum of the exponential values, and taking the negative logarithm to obtain the final loss value; the adjustment parameter is to adjust the distribution of the similarity to balance the contrast between positive and negative samples; Calculate the gradient of the loss value with respect to the parameters of the graph neural network model, and use the gradient descent algorithm to update the model parameters so that the similarity between complete positive sample pairs is maximized and the similarity between complete negative sample pairs is minimized to optimize the parameters in the graph neural network model.

Images