High-precision lightweight multi-process domain heterogeneous test data monitoring method and device
By selecting an adaptive evaluation model through data fusion, the problem of sensor evaluation accuracy in aircraft development was solved, achieving efficient sensor evaluation, improving the accuracy and efficiency of the aircraft development process, and reducing the consumption of computing resources.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHENGDU AIRCRAFT INDUSTRY GROUP
- Filing Date
- 2023-09-14
- Publication Date
- 2026-06-19
AI Technical Summary
In existing technologies, ground testing and evaluation methods during aircraft development are based on technical thresholds for each step of the process. This results in inaccuracies in sensor accuracy and an inability to effectively address the technical problems in the aircraft development process domain. Consequently, there is a large margin in the pass/fail criteria for sensors in each process domain, leading to unacceptable accuracy during flight and increasing development costs and time.
A multi-process domain heterogeneous test data monitoring method is adopted. After data fusion, different evaluation models are selected according to the number of features, including high-dimensional test data models based on kernel function statistics and low-dimensional test data models based on Gaussian processes. Combined with lightweight models, high-precision sensor evaluation is achieved.
By selecting an appropriate evaluation model after data fusion, the accuracy and efficiency of sensor ground testing and evaluation are improved, while the consumption of computing resources is reduced.
Smart Images
Figure CN117370153B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of aircraft condition monitoring technology, and more specifically to a high-precision, lightweight method and device for monitoring heterogeneous test data across multiple process domains. Background Technology
[0002] Before a test flight, an aircraft goes through various research and development processes, including design, manufacturing, installation, debugging, and ground testing. Before flight, each process requires extensive system-level testing to ensure that the aircraft meets the conditions for test flight, improve the efficiency of aircraft research and development, and avoid the loss of human and financial resources caused by test flight failures.
[0003] Currently, ground testing and evaluation methods are mainly based on a step-by-step process domain-based approach to determine the suitability of mission sensors. This involves evaluating each sensor from the design stage to determine if it meets the current process's compliance standards before proceeding to the next stage of testing. The compliance standards for each process are thresholds set based on expert experience. However, because these manually set thresholds have a large margin, even if a mission sensor meets the compliance standards in all process domains, it often fails to meet accuracy standards during actual flight. This not only reduces aircraft development and testing efficiency but also significantly increases development costs. Therefore, there is a need to establish a ground testing and evaluation model based on multi-process domain test data. During ground testing and evaluation, this model should utilize multi-process domain heterogeneous test data analysis to uncover deeper rules and predict whether the mission sensor's accuracy will meet standards. This has become a bottleneck limiting the accuracy of ground testing and evaluation and reducing development costs.
[0004] Due to the complexity of mission sensors, it is difficult to describe all data and the entire process using a single, universally applicable set of rules. Sensor data includes data from multiple process domains, such as assembly and takeoff. These multi-domain data exhibit different process areas and structural characteristics. Therefore, information from multiple process domains was fused before modeling. The dimensionality of the fused data is uncertain due to the characteristics of the sensors. Choosing a fixed model cannot accurately model based on dynamic dimensional data. If a machine learning model is built using high-dimensional data as input, the model becomes overly complex, computationally inefficient, and prone to overfitting and decreased accuracy with low-dimensional data. Conversely, if a model is trained using low-dimensional data as input, the model becomes too simple and cannot meet the accuracy requirements of higher-dimensional data input. Furthermore, traditional high-precision models are structurally complex, consume significant computational resources, and are time-consuming. Summary of the Invention
[0005] To address the problems and shortcomings of the existing technologies, this invention proposes a high-precision, lightweight, multi-process domain heterogeneous test data monitoring method and device. To reduce the consumption of model computational resources, this invention employs different methods to implement targeted modeling based on the size of the data feature dimensions, and the constructed model is a lightweight model, which greatly improves the accuracy of the ground test evaluation qualification judgment of aircraft mission sensors.
[0006] To achieve the above-mentioned objectives, the technical solution of the present invention is as follows:
[0007] A high-precision, lightweight method for monitoring heterogeneous test data across multiple process domains includes: firstly, acquiring multi-process domain aircraft mission sensor data and fusing the acquired data to obtain multi-process domain fused data; secondly, determining the dimensionality of the fused data based on the number of features, selecting a data evaluation model based on the dimensionality of the fused data, feeding the fused data into the evaluation model, and finally obtaining the predicted evaluation result; if the number of features after multi-process domain data fusion is large, a high-dimensional test data model based on kernel function statistics is selected as the evaluation model; if the number of features after multi-process domain data fusion is small, a low-dimensional test data model based on Gaussian processes is selected as the evaluation model.
[0008] Preferably, if the number of features after multi-process domain data fusion is large, a high-dimensional test data model based on kernel function statistics is selected as the evaluation model, including:
[0009] The input to the model training set is the fused data from the ground testing process and the flight test process, denoted as...
[0010]
[0011] in, Input for each sample; for A dimensional vector; the model's single-sample output is p represents the number of sensor test sessions. To determine the number of features in the fused data;
[0012] The optimization objective of the high-dimensional test data model can then be written as:
[0013]
[0014] Where w is the model weight, b is the bias term, ||w|| is the L2 norm of w, and l 0 / 1 It is expressed as follows:
[0015]
[0016] Equation (3) is optimized using the hinge loss function, resulting in the following:
[0017]
[0018] Introduce slack variable ξ into formula (4) i If ≥0, then:
[0019]
[0020] The Lagrange function is obtained using the Lagrange multiplier method:
[0021]
[0022] Where, α i ≥0,μ i ≥0 is a Lagrange multiplier, leading to its dual problem:
[0023]
[0024] Solve The optimal separating hyperplane, i.e., the desired model decision function, is obtained as follows:
[0025]
[0026] in, The RBF kernel function has the following expression:
[0027]
[0028] Where υ0 represents the hyperparameters to be initialized, represents the variance of the model; d is the dimension of the input data; ω1,ω2,...,ω d This represents the distance dimension, a hyperparameter to be initialized.
[0029] Preferably, if the number of features after multi-process domain data fusion is small, a low-dimensional test data model based on Gaussian processes is selected as the evaluation model, including:
[0030] The training set for the low-dimensional test data model is the same as that for the high-dimensional test data model, and the model feature input is... The output is p represents the number of sensor test sessions. To integrate the number of data features; the output of each function value Treat it as a random variable, its corresponding function value A set that constitutes a finite set of random variables;
[0031] Suppose that for a given finite input dataset, which follows a joint Gaussian distribution, then f(x) constitutes a GP:
[0032]
[0033] in, Represents any random input sample; It is a random variable The mean, Describing covariance, reflect as well as The correlation between them.
[0034] Preferably, the method further includes: comparing the evaluation prediction results obtained by the model with the actual results to measure the effectiveness of the model evaluation results on the test data.
[0035] Preferably, the model evaluation method using a confusion matrix is used to measure the effectiveness of the model evaluation results on the test data.
[0036] A high-precision, lightweight multi-process-domain heterogeneous test data monitoring device, the device being used to implement the above-described method, comprising:
[0037] The data acquisition module is used to acquire multi-process domain aircraft mission sensor data;
[0038] The data fusion module is used to fuse the collected data to obtain multi-process domain fused data;
[0039] The test data model selection module is used to determine the dimension of the fused data based on the number of features of the multi-process domain fused data, and select the test data evaluation model based on the dimension of the fused data.
[0040] The prediction output module evaluates the model based on the selected test data, performs prediction and evaluation on the multi-process domain fused data, and finally outputs the prediction and evaluation results.
[0041] A computer device includes a memory, a processor, and a computer program stored in the memory and executable in the processor, wherein when the processor executes the computer program, it implements the steps of the method described above.
[0042] A computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program, which, when executed in a computer processor, implements the steps of the method described above.
[0043] The beneficial effects of this invention are:
[0044] This invention can balance the requirements of lightweight model and high-precision data monitoring. By adaptively selecting the monitoring model through data dimensions, it performs well on data in multiple process domains and greatly improves the accuracy of ground test evaluation of aircraft mission sensors. Attached Figure Description
[0045] The foregoing and hereinafter detailed description of the invention becomes clearer when read in conjunction with the following drawings, in which:
[0046] Figure 1 This is a flowchart of the method of the present invention;
[0047] Figure 2 This is a structural diagram of the device of the present invention;
[0048] Figure 3 This is a schematic diagram illustrating the verification of the high-dimensional test data model evaluation results in Embodiment 2 of the present invention;
[0049] Figure 4 This is a schematic diagram illustrating the verification of the low-dimensional test data model evaluation results in Embodiment 2 of the present invention. Detailed Implementation
[0050] To enable those skilled in the art to better understand the technical solutions of this invention, several specific embodiments will be used to further illustrate the technical solutions for achieving the objectives of this invention. It should be noted that the technical solutions claimed by this invention include, but are not limited to, the following embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without inventive effort should fall within the scope of protection of this invention.
[0051] Example 1
[0052] Currently, ground testing and evaluation methods are mainly based on a step-by-step process domain-based approach to determine the suitability of mission sensors. This involves evaluating each sensor from the design stage to determine if it meets the current process's compliance standards before proceeding to the next stage of testing. The compliance standards for each process are thresholds set based on expert experience. However, because these manually set thresholds have a large margin, even if a mission sensor meets the compliance standards in all process domains, it often fails to meet accuracy standards during actual flight. This not only reduces aircraft development and testing efficiency but also significantly increases development costs. Therefore, there is a need to establish a ground testing and evaluation model based on multi-process domain test data. During ground testing and evaluation, this model should utilize multi-process domain heterogeneous test data analysis to uncover deeper rules and predict whether the mission sensor's accuracy will meet standards. This has become a bottleneck limiting the accuracy of ground testing and evaluation and reducing development costs.
[0053] Due to the complexity of mission sensors, it is difficult to describe all data and the entire process using a single, universally applicable set of rules. Sensor data includes data from multiple process domains, such as assembly and takeoff. These multi-domain data exhibit different process areas and structural characteristics. Therefore, information from multiple process domains was fused before modeling. The dimensionality of the fused data is uncertain due to the characteristics of the sensors. Choosing a fixed model cannot accurately model based on dynamic dimensional data. If a machine learning model is built using high-dimensional data as input, the model becomes overly complex, computationally inefficient, and prone to overfitting and decreased accuracy with low-dimensional data. Conversely, if a model is trained using low-dimensional data as input, the model becomes too simple and cannot meet the accuracy requirements of higher-dimensional data input. Furthermore, traditional high-precision models are structurally complex, consume significant computational resources, and are time-consuming.
[0054] Based on this, embodiments of the present invention propose a high-precision, lightweight, multi-process domain heterogeneous test data monitoring method and device. The method of the present invention can take into account both the lightweight requirements of the model and the accuracy requirements of data monitoring. By adaptively selecting the monitoring model through data dimensions, it performs well on multi-process domain data and greatly improves the accuracy of the qualification judgment of ground test evaluation of aircraft mission sensors.
[0055] This embodiment first discloses a high-precision, lightweight method for monitoring heterogeneous test data across multiple process domains, the method being as follows:
[0056] First, aircraft mission sensor data from multiple process domains are collected. Then, the collected data are fused to obtain multi-process domain fused data. Next, the dimensionality of the fused data is determined based on the number of features in the multi-process domain fused data. A data evaluation model is selected based on the dimensionality of the fused data, and the fused data is fed into the evaluation model to obtain the final evaluation result. If the number of features after multi-process domain data fusion is large, a high-dimensional test data model based on kernel function statistics is selected as the evaluation model. If the number of features after multi-process domain data fusion is small, a low-dimensional test data model based on Gaussian processes is selected as the evaluation model.
[0057] In this embodiment, it should be noted that the collected multi-process domain sensor data includes sensor data from processes such as flight testing and ground testing, where ground testing includes processes such as factory departure, incoming inspection, and installation. Furthermore, for multi-process domain data fusion, the fused data includes time-series data, frequency-domain sequence data, etc. Data with actual physical meaning from which data features can be extracted can be fused. Data fusion itself is a relatively conventional method in the field, so this invention will not elaborate on it further.
[0058] In this embodiment, it should also be noted that the number of data features reflects the size of the data dimension. The size of the fused data dimension is mainly determined by the shape of the fused data matrix, that is, the dimensionality of the data is determined by the shape of the fused data matrix. In this invention, if the dimension of the fused data exceeds 10, it is considered to have a large number of features and is high-dimensional data. A high-dimensional test data model based on kernel function statistics is selected as the evaluation model to evaluate the data. Conversely, if the dimension is less than 10, it is considered to have a small number of features, and a low-dimensional test data model based on Gaussian processes is selected as the evaluation model to evaluate the data.
[0059] Furthermore, in this embodiment, when the fused data features are large, indicating high-dimensional data, a high-dimensional test data model based on kernel function statistics is used as the evaluation model, and the final evaluation result is as follows:
[0060] The input to the model training set is the fused data from the ground testing process and the flight test process, denoted as...
[0061]
[0062] Let the input for each sample be denoted as for A dimensional vector, the single-sample output of the model is denoted as... This indicates whether the mission system is qualified for this sortie. 0 indicates an anomaly during the flight, and 1 indicates qualification. Here, p represents the sensor test sortie. The number of fusion features;
[0063] The optimization objective of the high-dimensional test data model can then be written as:
[0064]
[0065] Where w is the model weight, b is the bias term, ||w|| is the L2 norm of w, and l 0 / 1 It is expressed as follows:
[0066]
[0067] Since the 0 / 1 loss function is neither convex nor continuous, some convex continuous functions are used, and these functions are l. 0 / 1 The hinge loss function is replaced by the upper bound of the hinge loss function. Substituting the hinge loss function into equation (2), we get the following:
[0068]
[0069] Introduce slack variable ξ into formula (4) i ≥0, rewrite formula (4) as:
[0070]
[0071] The Lagrange function is obtained using the Lagrange multiplier method:
[0072]
[0073] Where, α i ≥0,μ i ≥0 is a Lagrange multiplier, leading to its dual problem:
[0074]
[0075] Solve The optimal separating hyperplane, i.e., the desired model decision function, is obtained as follows:
[0076]
[0077] in, The RBF kernel function has the following expression:
[0078]
[0079] Where υ0 represents the hyperparameters to be initialized, represents the variance of the model; d is the dimension of the input data; ω1,ω2,...,ω d This represents the distance dimension, a hyperparameter to be initialized. The dimension of the distance dimension should be consistent with the dimension of the input data. In practical applications, the initial value of the hyperparameter can be set to a random value between 0 and 1.
[0080] Furthermore, in this embodiment, when the fused data feature data is small, indicating that it is low-dimensional data, a low-dimensional test data model based on Gaussian process is used as the evaluation model, and the final evaluation result is as follows:
[0081] A stochastic process {X(t): t∈p} satisfies the following condition for any positive integer and any time interval: 0 < t1 < t2 < ... < t p Random variables X(t1), X(t2), ..., X(t) p If a stochastic process follows a joint normal distribution, then it is called a GP (Gross Probability Path), and its n-dimensional joint probability density is expressed as follows:
[0082]
[0083] Among them, u k Represents the random variable X(t) k The mathematical expectation of u k =E[X(t) k )];σ k It is a random variable X(t) kThe standard deviation of the normalized covariance matrix; |C| represents the determinant of the normalized covariance matrix, |C jk | is the element c in the determinant | C | jk The algebraic cofactor, c jk The normalized covariance function:
[0084]
[0085] In practical applications, the collected data consists of target values containing noise, as shown below:
[0086] y = f(x) + ε (12);
[0087] Where ε is the added white noise, which follows a pattern with mean 0 and variance σ. 2 The normal distribution, i.e., ε~N(0,σ) 2 For different x, the additional noise is independent and identically distributed.
[0088] The idea behind using Generalized Approach (GP) modeling is that it's unnecessary to definitively give f(x) a specific parameterized or non-parameterized expression; instead, it can be treated as a random variable. The training dataset for the model is the same as above, i.e., the model feature input is... The output is p represents the number of sensor test sessions. To integrate the number of data features, the output of each function value is... Treat it as a random variable, its corresponding function value A set of finite random variables. Suppose that for a given finite input dataset, which follows a joint Gaussian distribution, then f(x) constitutes a GP:
[0089]
[0090] in Represents any random input sample; It is a random variable The mean, Describing covariance, reflect as well as The correlation between them.
[0091] Regarding the concept of GP, when each input... Corresponding to t i At time t, f(x) represents the Gaussian process X(t). For the application scenario in this invention, using... Replace t if(x) replaces X(t). When this method is used for modeling, its properties are entirely determined by the mean function and the covariance function. While the mean and covariance functions can have different expressions in practical applications, data standardization is usually required beforehand, and the mean function can therefore be set to zero mean. In this case, the prior distribution of the GP is entirely determined by the covariance function and its included hyperparameters. The commonly used covariance function is the squared exponential covariance function, i.e., the RBF kernel function. Therefore, as a probabilistic classification model, this method can output the accuracy prediction results of target sorties, i.e., whether the mission system is qualified, and also obtain the confidence level of the prediction results.
[0092] Furthermore, after obtaining the evaluation results through the test data model, this implementation will also verify the validity of the evaluation results to ensure the reliability of the data. Specifically, the predicted evaluation results obtained from the test data model and the actual results are fed into an evaluation model based on a confusion matrix for data validity evaluation. The model evaluation method based on a confusion matrix is a commonly used technique for verifying model validity, and therefore will not be elaborated further.
[0093] Based on the same inventive concept, this invention also provides a high-precision, lightweight multi-process domain heterogeneous test data monitoring device. This device is used to implement the high-precision, lightweight multi-process domain heterogeneous test data monitoring described in the above embodiments, as illustrated in the following embodiments. As used below, the terms "unit" or "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated. Specifically, refer to the appendix to the specification. Figure 2 The device may include a data acquisition module 201, a data fusion module 202, a test data model selection module 203, and a prediction output module 204; wherein,
[0094] Data acquisition module 201 is used to acquire multi-process domain aircraft mission sensor data;
[0095] Data fusion module 202 is used to fuse the collected data to obtain multi-process domain fused data;
[0096] The test data model selection module 203 is used to determine the dimension of the fused data based on the number of features of the multi-process domain fused data, and select the test data evaluation model based on the dimension of the fused data.
[0097] The prediction output module 204 evaluates the multi-process domain fused data based on the selected test data evaluation model, and finally outputs the prediction evaluation results.
[0098] It should be noted that the systems, devices, models, or units described in the above embodiments can be implemented by computer chips or physical entities, or by products with certain functions. For ease of description, the above devices are described in this specification as various units based on their functions. Of course, in implementing this invention, the functions of each unit can be implemented in one or more software and / or hardware.
[0099] Furthermore, embodiments of the present invention also provide a computer device, which includes a memory, a processor, and a computer program stored in the memory and executable in the processor, wherein when the processor executes the computer program, it implements the steps of any of the above methods.
[0100] Furthermore, embodiments of the present invention also provide a computer-readable storage medium storing a computer program that, when executed in a computer processor, implements the steps of any of the above methods.
[0101] Example 2
[0102] In this case, flight parameters related to mission sensors are used as model inputs. Each parameter sample has a total length of 1000 sample points. Data from multiple process domains is fused and used as input to the monitoring data model. Different models are selected based on the different dimensions of the fused data to balance model accuracy and data complexity, thereby achieving high-precision modeling.
[0103] Table 2. List of UAV flight parameter data and units of each parameter.
[0104]
[0105] When the input data dimension is twelve, a high-dimensional test data evaluation model based on kernel function statistics is selected to model the flight data and determine whether the flight is qualified. The twelve-dimensional data fused from the six flight parameters in Table 2 is selected as the model input. The ground test evaluation model determines whether the mission system of each sortie is qualified based on the ground test accuracy data; the output is 0 when qualified and 1 when unqualified. A confusion matrix is used to measure the effectiveness of the method proposed in this embodiment. The model is trained using flight data from seven sorties. The effectiveness of the model on the data of entirely new sorties is obtained by referring to the appendix of the instruction manual. Figure 3 As shown.
[0106] When the input data dimension is six-dimensional, a low-dimensional test data evaluation model based on Gaussian processes is selected to model the flight data and determine whether the flight is qualified. The six-dimensional data, including altitude, pitch angle, and angle of attack from Table 2, is selected as the model input. The ground test evaluation model determines whether the mission system of each sortie is qualified based on the ground test accuracy data; the output is 0 when qualified and 1 when unqualified. A confusion matrix is used to measure the effectiveness of the proposed method. The model is trained using flight data from five sorties, and the effectiveness of the model is tested on data from entirely new sorties. The results are shown in the appendix to the specification. Figure 4 As shown.
[0107] The results above show that the proposed method can accurately determine the aircraft's state for both high-dimensional and low-dimensional data, providing strong support for the application of this invention.
[0108] The above description is merely a preferred embodiment of the present invention and is not intended to hinder the present invention in any way. Any simple modifications or equivalent changes made to the above embodiments based on the technical essence of the present invention shall fall within the protection scope of the present invention.
Claims
1. A high-precision, lightweight method for monitoring heterogeneous test data across multiple process domains, characterized in that, include: First, multi-process domain aircraft mission sensor data are collected, and the collected data are fused to obtain multi-process domain fused data. The dimension of the fused data is determined by the number of features of the multi-process domain fused data, a data evaluation model is selected based on the dimension of the fused data, the fused data is fed into the evaluation model, and finally the prediction evaluation result is obtained. If the number of features after fusing data from multiple process domains is large, a high-dimensional test data model based on kernel function statistics is selected as the evaluation model; if the number of features after fusing data from multiple process domains is small, a low-dimensional test data model based on Gaussian processes is selected as the evaluation model. When the dimension of the fused data exceeds 10, it indicates a large number of features; conversely, a smaller number of features indicates a smaller number of features.
2. The high-precision, lightweight, multi-process-domain heterogeneous test data monitoring method according to claim 1, characterized in that, If the number of features after fusing multi-process domain data is large, a high-dimensional test data model based on kernel function statistics is selected as the evaluation model, including: The input to the model training set is the fused data from the ground testing process and the flight test process, denoted as . （1）； in, Input for each sample; 1* A dimensional vector; the model's single-sample output is ; For sensor test sessions, To determine the number of features in the fused data; The optimization objective of the high-dimensional test data model can then be written as: （2）； in, For the model weights, For bias terms, for of Norm, It is expressed as follows: （3）； Equation (3) is optimized using the hinge loss function, resulting in the following: （4）； Introduce slack variables in formula (4) Then we have: （5）； The Lagrange function is obtained using the Lagrange multiplier method: （6）； in, These are Lagrange multipliers, leading to their dual problem: （7）； Solve The optimal separating hyperplane, i.e., the desired model decision function, is obtained as follows: （8）； in, The RBF kernel function has the following expression: （9）； in, This represents the hyperparameters to be initialized, and this represents the variance of the model. It is the dimension of the input data; This represents the distance dimension, a hyperparameter to be initialized.
3. The high-precision, lightweight multi-process domain heterogeneous test data monitoring method according to claim 1, characterized in that, If the number of features after fusing multi-process domain data is small, a low-dimensional test data model based on Gaussian processes is selected as the evaluation model, including: The training set for the low-dimensional test data model is the same as that for the high-dimensional test data model, and the model feature input is... The output is , For sensor test sessions, To integrate the number of data features; the output of each function value Treat it as a random variable, its corresponding function value A set that constitutes a finite set of random variables; Suppose that for a given finite input dataset, which follows a joint Gaussian distribution, then This constitutes a GP: （13）； in, , Represents any random input sample; It is a random variable The mean, Describing covariance, reflect as well as The correlation between them.
4. The high-precision, lightweight multi-process-domain heterogeneous test data monitoring method according to claim 1, characterized in that, The method further includes: comparing the evaluation prediction results obtained by the model with the actual results to measure the effectiveness of the model evaluation results on the test data.
5. The high-precision, lightweight, multi-process-domain heterogeneous test data monitoring method according to claim 4, characterized in that, The confusion matrix model evaluation method is used to measure the effectiveness of the model evaluation results on the test data.
6. A high-precision, lightweight multi-process-domain heterogeneous test data monitoring device, the device being used to implement the method described in any one of claims 1-5, characterized in that, include: The data acquisition module is used to acquire multi-process domain aircraft mission sensor data; The data fusion module is used to fuse the collected data to obtain multi-process domain fused data; The test data model selection module is used to determine the dimension of the fused data based on the number of features of the multi-process domain fused data, and select the test data evaluation model based on the dimension of the fused data. The prediction output module evaluates the model based on the selected test data, performs prediction and evaluation on the multi-process domain fused data, and finally outputs the prediction and evaluation results.
7. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable in the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method described in any one of claims 1-5.
8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed in a computer processor, implements the method described in any one of claims 1-5.