A dual-parameter adaptive enhancement method for cardiovascular disease diagnosis

Abstract

The present invention discloses a dual-parameter adjustment adaptive enhancement method for diagnosing cardiovascular diseases, which belongs to the field of integrated algorithms and machine learning. The method comprises inputting acquired diagnostic sample data into an optimal classifier of a pre-acquired adaptive enhancement model to obtain diagnostic reference information; wherein the method for obtaining the optimal classifier comprises: training a weak classifier based on the acquired training data; constructing an adaptive enhancement model of adjustable parameter pairs based on the trained weak classifier; iteratively optimizing the adaptive enhancement model using an improved Bayesian optimization method for parameter optimization to obtain the optimal classifier of the adaptive enhancement model. The present invention significantly improves the generalization and robustness of the model by introducing adjustable parameter pairs and combining parameter optimization at the same time, realizes automated and efficient processing of diagnostic sample data, provides doctors with more objective and accurate diagnostic reference information, and effectively improves the efficiency and accuracy of disease diagnosis.

Description

Technical Field

[0001] The invention relates to a dual-parameter adjustment adaptive enhancement method for cardiovascular disease diagnosis, and belongs to the field of integrated algorithms and machine learning. Background Art

[0002] Machine learning plays a vital role in artificial intelligence, building models through algorithms for data prediction and decision making. Classification is one of its core tasks, which is to assign samples to different categories. Ensemble learning methods, such as bagging and boosting, can effectively improve classification performance, especially when using unstable learning algorithms. In recent years, ensemble learning methods have been widely used in disease diagnosis, signal detection, time series prediction, and image classification.

[0003] Cardiovascular disease (CVD) has always been one of the top ten causes of death in the world, but the inconsistency and redundancy of clinical data have led to great challenges in diagnostic algorithms. So far, researchers have proposed a variety of dual-parameter adjustment adaptive enhancement algorithms for cardiovascular disease diagnosis, among which the most classic is the AdaBoost algorithm. However, there are some problems in the improvement of the AdaBoost algorithm in recent years: first, many methods only replace weak classifiers without improving the integrated framework of AdaBoost; second, some methods change the weight classification of samples without sufficient theoretical support; and the traditional AdaBoost algorithm, when used for cardiovascular diagnosis, faces the problem of low generalization and robustness of the algorithm when facing unbalanced and noisy data. Summary of the invention

[0004] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a dual-parameter adjustment adaptive enhancement method for the diagnosis of cardiovascular diseases, which realizes the automated and efficient processing of diagnostic sample data, provides doctors with more objective and accurate diagnostic reference information, and effectively improves the efficiency and accuracy of disease diagnosis.

[0005] To achieve the above object, the present invention is implemented by adopting the following technical solutions:

[0006] In a first aspect, the present invention provides a dual-parameter adjustment adaptive enhancement method for cardiovascular disease diagnosis, comprising:

[0007] Inputting the acquired diagnostic sample data into the optimal classifier of the pre-acquired adaptive enhancement model to obtain diagnostic reference information;

[0008] The method for obtaining the optimal classifier includes:

[0009] Based on the acquired training data, train a weak classifier;

[0010] Based on the trained weak classifier, an adaptive boosting model with adjustable parameter pairs is constructed;

[0011] The adaptive boosting model is iteratively optimized using an improved Bayesian optimization method for parameter optimization to obtain the optimal classifier of the adaptive boosting model.

[0012] Furthermore, the method also includes: preprocessing the acquired training data, including processing missing values, detecting outliers, data normalization and feature selection.

[0013] Furthermore, the training of the weak classifier based on the acquired training data includes:

[0014] Input preprocessed training data :

[0015] ;

[0016] in, Representative The feature vector of the samples, represent Tags, ; represents the total number of samples;

[0017] Initialize the weights of the training data using the following formula:

[0018] ;

[0019] ;

[0020] in refers to the overall sample weight distribution, The first round of iteration The weights assigned to each sample;

[0021] make ,in Indicates the sequence number of the current iteration. Refers to the total number of weak classifiers, and corresponds to the total number of iterations; The training dataset with weights is used to train weak classifiers , defined as follows:

[0022] ;

[0023] The weak classifier transforms the feature space The samples in the space are mapped to the label space ,in, Used to indicate whether a patient is diagnosed with cardiovascular disease. Indicates negative, Indicates positive.

[0024] Furthermore, the adaptive enhancement model of adjustable parameter pairs is constructed based on the trained weak classifier, including:

[0025] Compute the trained weak classifier The classification error rate on the training data set is as follows:

[0026] ;

[0027] In the formula, Indicates The feature vector of samples; Indicates The labels of samples, represents the classification error rate, It is The first iteration The weights assigned to each sample; is an indicator function, defined as follows:

[0028] ;

[0029] According to the classification error rate, calculate the weak classifier Weight , the formula is as follows:

[0030] ;

[0031] in, is the overall adjustment parameter for the weight update strength, Fine-tune parameters for update intensity;

[0032] According to the weak classifier weights, update the weight distribution of the training data set, The weight distribution of the training data set in the iteration and In the iteration The weight of the samples The formula is as follows:

[0033] ;

[0034] in It is In the iteration The weight of the samples;

[0035] ;

[0036] in is the normalization factor, defined as follows:

[0037] ;

[0038] The weighted majority voting method is used to combine multiple base models into an adaptive enhanced model with adjustable parameter pairs. The formula is as follows:

[0039] ;

[0040] ;

[0041] In the formula, is a linear combination of weighted weak classifiers, Represents an adaptive boosting model of adjustable parameter pairs.

[0042] Furthermore, the adaptive enhancement model is iteratively optimized using the improved Bayesian optimization method for parameter optimization to obtain the optimal classifier of the adaptive enhancement model, including:

[0043] Initialize the parameter observation set using Latin Hypercube sampling , the formula is as follows:

[0044] ;

[0045] in is the current parameter combination, is independent and identically distributed Gaussian noise, It means that the objective function is The value of the parameter combination;

[0046] Use Gaussian process to analyze the parameter observation set The observed data in is modeled to obtain a Gaussian process model;

[0047] The mean of the Gaussian process model and covariance function The formula is as follows:

[0048] ;

[0049] in Represents the variance of the observation noise; it is the variance parameter of Gaussian noise and is used to represent the uncertainty in the observation data; Represents the identity matrix, which is used to regularize the kernel matrix together with the noise variance to ensure stable calculation when the matrix is ​​inverted. is the covariance matrix, represents the target value set of the observed data, is the covariance function in the Gaussian process, which is used to represent the observation point currently considered and observation points The correlation between

[0050] ;

[0051] in, is the observation point currently considered, represents the predicted target point, and are the sets of observation points and objective function values, respectively. is the covariance function in the Gaussian process, used to represent the observation points and the predicted target point The correlation between

[0052] Optimizing Hyperparameters of Gaussian Process Models Using Improved Markov Chain Monte Carlo Method , the posterior distribution It is expressed as:

[0053] ;

[0054] Represents a given Gaussian process hyperparameter The likelihood function under the condition is Represents the hyperparameters of the Gaussian process The prior distribution of

[0055] Among them, the sampling from the posterior distribution is done by the Metropolis-Hastings algorithm, and the sampling acceptance rate Defined as:

[0056] ;

[0057] in, represents the candidate new state generated by the current step, represents the improved Markov chain Monte Carlo method to optimize the hyperparameters of the Gaussian process model, Represents the hyperparameters after a given optimization The likelihood function in the case of Represents the hyperparameters of the Gaussian process The prior distribution of

[0058] Select the next evaluation point using the improved Thompson sampling method , the sampling formula is:

[0059] ;

[0060] in, is the probability distribution of Thompson sampling, For the current observation data set, the expected improvement of the acquisition function is expressed as:

[0061] ;

[0062] in, Indicated in The objective function value at is the point with the best performance in the current observation set;

[0063] The formula for updating the observation set is as follows:

[0064] ;

[0065] in, represents the observation noise;

[0066] Retrain the Gaussian process model until the convergence condition is met and the optimal weight value of the adaptive enhancement model is obtained. and , and train the optimal classifier at the same time. The formula is as follows:

[0067] ;

[0068] in, Represents the final optimal classifier based on the integration of a series of weighted weak classifiers. To adjust the optimal parameters for model performance, represents a weighted linear combination of multiple weak classifiers, Indicates The weights of weak classifiers, Indicates A weak classifier.

[0069] In a second aspect, the present invention provides a dual-parameter adjustment adaptive enhancement device for cardiovascular disease diagnosis, comprising:

[0070] A diagnosis module, used to input the acquired diagnosis sample data into the optimal classifier of the pre-acquired adaptive enhancement model to obtain diagnosis reference information;

[0071] The method for obtaining the optimal classifier includes:

[0072] Based on the acquired training data, train a weak classifier;

[0073] Based on the trained weak classifier, an adaptive boosting model with adjustable parameter pairs is constructed;

[0074] The adaptive boosting model is iteratively optimized using an improved Bayesian optimization method for parameter optimization to obtain the optimal classifier of the adaptive boosting model.

[0075] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the aforementioned methods.

[0076] In a fourth aspect, the present invention provides a computer device, comprising:

[0077] Memory, for storing computer programs / instructions;

[0078] A processor is used to execute the computer program / instructions to implement the steps of any of the aforementioned methods.

[0079] In a fifth aspect, the present invention provides a computer program product, comprising a computer program / instruction, which implements the steps of any of the aforementioned methods when executed by a processor.

[0080] Compared with the prior art, the present invention has the following beneficial effects:

[0081] 1. The present invention provides a dual-parameter adjustment adaptive enhancement method for cardiovascular disease diagnosis. By introducing adjustable parameter pairs and combining parameter optimization, the generalization and robustness of the model are significantly improved, and the automated and efficient processing of diagnostic sample data is realized, providing doctors with more objective and accurate diagnostic reference information, effectively improving the efficiency and accuracy of disease diagnosis;

[0082] 2. The present invention provides a dual-parameter adjustment adaptive enhancement method for the diagnosis of cardiovascular diseases. By improving the traditional Bayesian optimization algorithm, the efficiency and accuracy of parameter optimization are improved, thereby improving the usability of the model. The model can optimize parameters autonomously and has significant convenience and practicality. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] Figure 1 It is a full flow chart of a dual-parameter adjustment adaptive enhancement method for cardiovascular disease diagnosis provided by an embodiment of the present invention;

[0084] Figure 2 It is a flowchart of an algorithm implementation of a dual-parameter adjustment adaptive enhancement method for cardiovascular disease diagnosis provided by an embodiment of the present invention;

[0085] Figure 3 The present invention provides a flowchart of data preprocessing of a dual-parameter adjustment adaptive enhancement method for diagnosing cardiovascular diseases. DETAILED DESCRIPTION

[0086] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0087] Embodiment 1: This embodiment introduces a dual-parameter adjustment adaptive enhancement method for cardiovascular disease diagnosis, comprising:

[0088] Inputting the acquired diagnostic sample data into the optimal classifier of the pre-acquired adaptive enhancement model to obtain diagnostic reference information;

[0089] The method for obtaining the optimal classifier includes:

[0090] Based on the acquired training data, train a weak classifier;

[0091] Based on the trained weak classifier, an adaptive boosting model with adjustable parameter pairs is constructed;

[0092] The adaptive boosting model is iteratively optimized using an improved Bayesian optimization method for parameter optimization to obtain the optimal classifier of the adaptive boosting model.

[0093] The dual-parameter adjustment adaptive enhancement method for cardiovascular disease diagnosis provided in this embodiment is intended to solve the problem of low generalization and robustness of the traditional AdaBoost algorithm when facing unbalanced and noisy data when used for cardiovascular diagnosis. This method introduces an improved weight adjustment strategy, and dynamically optimizes and adjusts the adaptive algorithm after the integrated framework, and dynamically selects the weight distribution of the integrated algorithm according to the data sample, thereby effectively avoiding the problem of poor classification effect of the AdaBoost algorithm when facing a small data sample and redundant data. The application process specifically involves the following steps:

[0094] Step 1: Prepare data for diagnosis and perform data preprocessing.

[0095] The data preprocessing steps include handling missing values, checking outliers, and data normalization. Since outliers in the medical field may have important clinical significance, we use the improved quartile method to detect outliers. In this method, the internal limit is defined as (upper quartile) plus (interquartile range) and (lower quartile) minus ; while the outer limit is defined as and Data points outside the inner limits are considered “mild” outliers, meaning they deviate from the normal range but are not extreme; data points outside the outer limits are considered “serious” outliers, meaning they deviate significantly from the normal distribution and may have an impact on the overall results.

[0096] For data identified as "serious" outliers, before deciding whether to process them, careful consideration and expert evaluation are required to determine the treatment method. Finally, the pre-processed data is obtained.

[0097] Step 2: Build an adaptive enhancement model with adjustable parameters (APPBoost). The specific steps are as follows:

[0098] Input: preprocessed data ;

[0099] in, Indicates The feature vector of a sample contains multiple feature values ​​of the sample. ,in is the feature space, containing all possible feature combinations, express dimensional real number space. Indicates The labels of samples, , used to indicate whether the patient is diagnosed with the cardiovascular disease, is a specific tag, Indicates negative, Indicates positive. middle, Representative The feature vector of the samples, Represents its label;

[0100] Parameter pair , weak classifier ;

[0101] Output: Final classifier .

[0102] (1) Initialize the weights of the training data:

[0103] ;

[0104] in refers to the overall sample weight distribution, The first round of iteration N The weights assigned to the samples.

[0105] (2) ,in Indicates the sequence number of the current iteration. Refers to the total number of weak classifiers, which also corresponds to the total number of iterations :

[0106] (a) Use The training dataset with weights is used to train weak classifiers :

[0107] ;

[0108] The weak classifier transforms the feature space The samples in the space are mapped to the label space .

[0109] (b) Calculation Classification error rate on the training dataset:

[0110] ;

[0111] In the formula, represents the classification error rate, It is The first iteration The weights assigned to each sample; is an indicator function, which is defined as follows:

[0112] .

[0113] (c) Calculate weak classifiers Weight :

[0114] ;

[0115] in, and is a pair of weight adjustment parameters.

[0116] (d) Update the weight distribution of the training dataset. The weight distribution of the training data set in the iteration and In the iteration The weight of the samples The formula is as follows:

[0117] ;

[0118] in It is In the iteration The weight of the samples;

[0119] ;

[0120] in is the normalization factor, defined as follows:

[0121] .

[0122] (3) Combined model: final output model , that is, combining multiple base models into a strong model through weighted majority voting:

[0123] ;

[0124] ;

[0125] In the formula, is a linear combination of weighted weak classifiers, Represents an adaptive boosting model of adjustable parameter pairs.

[0126] Step 3: Use improved Bayesian optimization for parameter optimization;

[0127] (1) Initialize the parameter observation set using the Latin Hypercube sampling method :

[0128] ;

[0129] in is the current parameter combination, is independent and identically distributed Gaussian noise.

[0130] Subsequently, the observed data are modeled using Gaussian process to obtain the Gaussian process model;

[0131] Mean of the Gaussian Process Model and covariance function :

[0132] ;

[0133] in Represents the variance of the observation noise, which is the variance parameter of Gaussian noise and is used to represent the uncertainty in the observation data. Represents the identity matrix, which is used to regularize the kernel matrix together with the noise variance to ensure stable calculation when inverting the matrix;

[0134] ;

[0135] in, is the observation point currently considered, represents the predicted target point, and are the sets of observation points and objective function values, respectively. is the covariance matrix.

[0136] (2) Using the improved Markov chain Monte Carlo method to optimize GP hyperparameters , the posterior distribution It is expressed as:

[0137] ;

[0138] Among them, the sampling from the posterior distribution is done by the Metropolis-Hastings algorithm, and the sampling acceptance rate is Defined as:

[0139] .

[0140] (3) Collection function

[0141] Select the next evaluation point using the improved Thompson sampling method (Thompson) , the sampling formula is:

[0142] ;

[0143] in, is the probability distribution of Thompson sampling, For the current observation data set, the expected improvement of the acquisition function is expressed as:

[0144] ;

[0145] in, Indicated in The objective function value at It is the point with the best performance in the current observation set.

[0146] (4) Iterative update

[0147] In each iteration, the observation set is updated and the GP model is retrained until convergence.

[0148] Update the observation set:

[0149] ;

[0150] Retrain the GP model and repeat steps (2) and (3) until the convergence condition is met (for example, the number of evaluations reaches a preset upper limit or the improvement is lower than a threshold).

[0151] Step 4: Obtain the optimal adjustable parameter weight value for the adaptive boosting algorithm (APPBoost) through Bayesian optimization and , and train the optimal final classifier at the same time:

[0152] .

[0153] Step 5: Input the required diagnostic sample data into the trained optimal final classifier for diagnosis.

[0154] The contents involved in the above embodiment are described below in conjunction with a preferred embodiment.

[0155] like Figure 1 As shown, the required training data is preprocessed first. The specific data processing process is as follows Figure 3 shown.

[0156] (1) The first is the dataset preparation stage. If a directly generated simulated dataset is used for experiments, no further processing is required. However, when testing on a medical dataset, preprocessing is required to obtain the best prediction model, including handling missing values, detecting outliers, data normalization, and feature selection.

[0157] (2) Next is outlier detection. In the medical field, outliers may have important clinical significance. The present invention uses an improved quartile method to detect outliers. Specifically, the internal limit is defined as the upper quartile. Interquartile range and lower quartile ; The outer limit is defined as and ; Data points outside the inner limits are considered “mild” anomalies, indicating deviations from the normal range but not extreme; Data points outside the outer limits are considered “serious” anomalies, indicating significant deviations from the normal distribution and may affect the overall results.

[0158] (3) The next step is to deal with outliers. For data points identified as "serious" anomalies, we will decide whether to deal with them after careful consideration and consultation with relevant experts. Finally, we get the processed data. .

[0159] 2. Construct an adaptive enhancement model (APBoost) with adjustable parameters. The specific steps are as follows:

[0160] Input: preprocessed data ;

[0161] in, Indicates The feature vector of a sample contains multiple feature values ​​of the sample. ,in is the feature space, containing all possible feature combinations, express dimensional real number space. Indicates The labels of samples, , used to indicate whether the patient is diagnosed with the cardiovascular disease, is a specific tag, Indicates negative, Indicates positive. middle, Representative The feature vector of the samples, Represents its label;

[0162] Parameter pair , weak classifier ;

[0163] Output: Final classifier .

[0164] In actual use, weak classifiers can be used In this example, we choose decision tree stump as weak classifier A decision stump is a decision tree with a depth of 1 that selects a feature and a threshold , the data set is divided into two parts, each corresponding to a different prediction label. Specifically, given the input data and its corresponding weight , the decision tree constructs the optimal classification rule by traversing all features and possible threshold combinations and selecting the combination that can minimize the weighted classification error rate. Finally, the weak classifier outputs a prediction value , which divides the input samples into two categories and then integrates with other weak classifiers to form the final strong classifier .

[0165] (1) Figure 2 As shown, initialize the weights of the training data:

[0166] ;

[0167] in refers to the overall sample weight distribution, The first round of iteration N The weights assigned to the samples.

[0168] (2) ,in Indicates the sequence number of the current iteration. Refers to the total number of weak classifiers, which also corresponds to the total number of iterations :

[0169] (a) Use The training dataset with weights is used to train weak classifiers :

[0170] ;

[0171] The weak classifier transforms the feature space The samples in the space are mapped to the label space .

[0172] (b) Calculation Classification error rate on the training dataset :

[0173] ;

[0174] In the formula, represents the classification error rate, It is The first iteration The weights assigned to each sample; is an indicator function, which is defined as follows:

[0175] .

[0176] (c) Calculate weak classifiers Weight , the formula is as follows:

[0177] ;

[0178] in, and is a pair of weight adjustment parameters.

[0179] (d) Update the weight distribution of the training dataset. The weight distribution of the training data set in the iteration and In the iteration The weight of the samples The formula is as follows:

[0180] ;

[0181] in It is In the iteration The weight of the samples;

[0182] ;

[0183] in is the normalization factor, defined as follows:

[0184] .

[0185] (3) Combined model: final output model , that is, combining multiple base models into a strong model through weighted majority voting:

[0186] ;

[0187] ;

[0188] In the formula, is a linear combination of weighted weak classifiers, Represents an adaptive boosting model of adjustable parameter pairs.

[0189] 3. The average model performance evaluation value is used as the objective function, and the improved Bayesian optimization method is used to optimize the adjustment parameters:

[0190] The present invention uses indicators such as accuracy, precision, recall, F1 score, and area under the receiver operating characteristic (ROC) curve (AUC) to evaluate the performance of the classification model.

[0191] Table 1 Confusion matrix

[0192]

[0193] Based on the binary confusion matrix in Table 1, the calculation formula of the evaluation index is as follows.

[0194] Accuracy It reflects the overall performance of the model and indicates the proportion of correctly predicted instances to the total predicted instances.

[0195] .

[0196] Accuracy It is an indicator used to evaluate the performance of the classification model, measuring the proportion of true samples among the samples predicted to be positive samples.

[0197] .

[0198] Recall It indicates the ability of the model to identify target instances in the dataset, that is, how many positive samples are predicted as positive samples.

[0199] .

[0200] F1 score is a way to summarize precision and recall, representing their harmonic mean.

[0201] .

[0202] The ROC curve is a plot with the true positive rate (TPR) as the vertical axis and the false positive rate (FPR) as the horizontal axis, which usually changes with the change of the threshold. When the threshold is set to the maximum or minimum, the model predicts positive or negative, and the TPR and FPR are both 1 or 0. The AUC can be obtained by summing the area under the ROC curve. A higher AUC indicates better classification model performance.

[0203] For example, when diagnosing cardiovascular disease, the present invention selects accuracy as the main evaluation index, while taking other evaluation indexes into consideration;

[0204] ;

[0205] in They are the weight parameters of the corresponding evaluation indicators in the final evaluation value, which can be further adjusted according to the specific situation. Parameter combination ;

[0206] The present invention then uses an improved Bayesian optimization method for optimization:

[0207] (1) Initialize the parameter observation set using the Latin Hypercube Sampling (LHS) method :

[0208] ;

[0209] in is the current parameter combination, is independent and identically distributed Gaussian noise.

[0210] Use Gaussian process to model the observed data and obtain Gaussian process model;

[0211] Mean of the Gaussian Process Model and covariance function :

[0212] ;

[0213] in Represents the variance of the observation noise, which is the variance parameter of Gaussian noise and is used to represent the uncertainty in the observation data. Represents the identity matrix, which is used to regularize the kernel matrix together with the noise variance to ensure stable calculation when inverting the matrix;

[0214] ;

[0215] in, is the observation point currently considered, represents the predicted target point, and are the sets of observation points and objective function values, respectively. is the covariance matrix.

[0216] (2) Using the improved Markov chain Monte Carlo method to optimize GP hyperparameters , the posterior distribution It is expressed as:

[0217] ;

[0218] Among them, the sampling from the posterior distribution is done by the Metropolis-Hastings algorithm, and the sampling acceptance rate is Defined as:

[0219] .

[0220] (3) Collection function

[0221] Select the next evaluation point using the improved Thompson sampling method (Thompson) , the sampling formula is:

[0222] ;

[0223] in, is the probability distribution of Thompson sampling, For the current observation data set, the expected improvement of the acquisition function is expressed as:

[0224] ;

[0225] in, Indicated in The objective function value at It is the point with the best performance in the current observation set.

[0226] (4) Iterative update

[0227] In each iteration, the observation set is updated and the GP model is retrained until convergence.

[0228] Update the observation set:

[0229] ;

[0230] Retrain the GP model and repeat steps 2 and 3 until the convergence condition is met (such as the number of evaluations reaches the preset upper limit or the improvement is lower than the threshold).

[0231] Finally, the optimal APBoost algorithm weight value is obtained through Bayesian optimization and , and train the optimal final classifier at the same time:

[0232] .

[0233] 4. Input the required diagnostic sample data into the trained optimal final classifier for diagnosis.

[0234] Enter data for diagnostics , which contains multiple possible characteristics of the patient related to cardiovascular disease.

[0235] Output diagnostic results , and the diagnosis results, Indicates negative, Indicates positive.

[0236] The present invention proposes an improved method (APPBoost) of the traditional AdaBoost algorithm for the diagnosis of cardiovascular diseases, which effectively solves many deficiencies in traditional classification technology. Through experiments on a real cardiovascular disease data set, the results show that the classification accuracy of the traditional AdaBoost algorithm on the data set is 82.1%, while the APPBoost algorithm of the present invention achieves a classification accuracy of 87.8%, significantly improving the classification performance. At the same time, the accuracy of the present invention is higher than that of commonly used machine learning and deep learning methods such as random forest (83.7%), XGBoost (80.2%), CNN (80.7%), and LSTM (83.5%).

[0237] By introducing an improved weight adjustment mechanism, the present invention can better handle noise and complex sample distribution in classification problems, thereby improving overall classification accuracy. The present invention performs well in multiple evaluation indicators, including accuracy, precision, recall, F1 score, and area under the ROC curve (AUC), with the recall rate being particularly outstanding, reaching 84.9%.

[0238] The method of the present invention shows great application potential in the dual-parameter adaptive enhancement for cardiovascular disease diagnosis. The dual-parameter adaptive enhancement for cardiovascular disease diagnosis requires high accuracy and high recall to ensure accurate identification of patients with the disease and reduce the missed diagnosis rate. The present invention can significantly improve the classification model's ability to identify patients with cardiovascular disease and reduce the risk of misdiagnosis and missed diagnosis.

[0239] The present invention uses an improved Bayesian optimization method to automatically adjust parameters to maximize the classification effect. Compared with the traditional grid search method, Bayesian optimization significantly reduces the time cost of parameter tuning, improves the overall computational efficiency of the algorithm, and further enhances its practicality in clinical diagnosis applications.

[0240] Embodiment 2: This embodiment provides a dual-parameter adjustment adaptive enhancement device for cardiovascular disease diagnosis, comprising:

[0241] A diagnosis module, used to input the acquired diagnosis sample data into the optimal classifier of the pre-acquired adaptive enhancement model to obtain diagnosis reference information;

[0242] The method for obtaining the optimal classifier includes:

[0243] Based on the acquired training data, train a weak classifier;

[0244] Based on the trained weak classifier, an adaptive boosting model with adjustable parameter pairs is constructed;

[0245] The adaptive boosting model is iteratively optimized using an improved Bayesian optimization method for parameter optimization to obtain the optimal classifier of the adaptive boosting model.

[0246] The specific functional implementation of each of the above modules can be found in the relevant contents of the method in Example 1 and will not be elaborated here.

[0247] Embodiment 3, this embodiment provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the steps of any one of the methods described in Embodiment 1 are implemented.

[0248] Embodiment 4: This embodiment provides a computer device, including:

[0249] Memory, for storing computer programs / instructions;

[0250] A processor, configured to execute the computer program / instructions to implement the steps of any one of the methods described in Example 1.

[0251] Embodiment 5: This embodiment provides a computer program product, including a computer program / instruction, which implements the steps of any method described in Embodiment 1 when executed by a processor.

[0252] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

[0253] It should be understood by those skilled in the art that the embodiments of the present disclosure may be provided as methods, systems or computer program products. Therefore, the present disclosure may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the present disclosure may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0254] The present disclosure is described with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present disclosure. It should be understood that each process and / or block in the flowchart and / or block diagram and the combination of processes and / or blocks in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0255] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0256] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0257] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present disclosure rather than to limit its protection scope. Although the present disclosure has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that after reading the present disclosure, those skilled in the art can still make various changes, modifications or equivalent substitutions to the specific implementation methods of the invention, but these changes, modifications or equivalent substitutions are all within the protection scope of the disclosed claims to be approved.

Claims

1. A dual-parameter adjustment adaptive enhancement method for cardiovascular disease diagnosis, characterized in that: include: Inputting the acquired diagnostic sample data into the optimal classifier of the pre-acquired adaptive enhancement model to obtain diagnostic reference information; The method for obtaining the optimal classifier includes: Based on the acquired training data, train a weak classifier; Based on the trained weak classifier, an adaptive boosting model with adjustable parameter pairs is constructed; The adaptive boosting model is iteratively optimized using an improved Bayesian optimization method for parameter optimization to obtain the optimal classifier of the adaptive boosting model. The step of training a weak classifier based on the acquired training data includes: Input preprocessed training data : ； in, Representative The feature vector of the samples, represent Tags, ; represents the total number of samples; Initialize the weights of the training data using the following formula: ； ； in refers to the overall sample weight distribution, The first round of iteration The weights assigned to each sample; make ,in Indicates the sequence number of the current iteration. Refers to the total number of weak classifiers, and corresponds to the total number of iterations; The training dataset with weights is used to train weak classifiers , defined as follows: ； The weak classifier transforms the feature space The samples in the space are mapped to the label space ,in, Used to indicate whether a patient is diagnosed with cardiovascular disease. Indicates negative, Indicates positive; The method of constructing an adaptive enhancement model with adjustable parameter pairs based on the trained weak classifier includes: Compute the trained weak classifier The classification error rate on the training data set is as follows: ； In the formula, Indicates The feature vector of samples; Indicates The labels of samples, represents the classification error rate, It is The first iteration The weights assigned to each sample; is an indicator function, defined as follows: ； According to the classification error rate, calculate the weak classifier Weight , the formula is as follows: ； in, is the overall adjustment parameter for the weight update strength, Fine-tune parameters for update intensity; According to the weak classifier weights, update the weight distribution of the training data set, The weight distribution of the training data set in the iteration and In the iteration The weight of the samples The formula is as follows: ； in It is In the iteration The weight of the samples; ； in is the normalization factor, defined as follows: ； The weighted majority voting method is used to combine multiple base models into an adaptive enhanced model with adjustable parameter pairs. The formula is as follows: ； ； In the formula, is a linear combination of weighted weak classifiers, An adaptive boosting model representing a pair of adjustable parameters; The method of iteratively optimizing the adaptive enhancement model using the improved Bayesian optimization method for parameter optimization to obtain the optimal classifier of the adaptive enhancement model includes: Initialize the parameter observation set using Latin Hypercube sampling , the formula is as follows: ； in is the current parameter combination, is independent and identically distributed Gaussian noise, It means that the objective function is The value of the parameter combination; Use Gaussian process to train the parameter observation set The observed data in is modeled to obtain a Gaussian process model; The mean of the Gaussian process model and covariance function The formula is as follows: ； in Represents the variance of the observation noise; it is the variance parameter of Gaussian noise and is used to represent the uncertainty in the observation data; Represents the identity matrix, which is used to regularize the kernel matrix together with the noise variance to ensure stable calculation when the matrix is ​​inverted. is the covariance matrix, represents the target value set of the observed data, is the covariance function in the Gaussian process, which is used to represent the observation point currently considered and observation points The correlation between ； in, is the observation point currently considered, represents the predicted target point, and are the sets of observation points and objective function values, respectively. is the covariance function in the Gaussian process, used to represent the observation points and the predicted target point The correlation between Optimizing Hyperparameters of Gaussian Process Models Using Improved Markov Chain Monte Carlo Method , the posterior distribution It is expressed as: ； Represents a given Gaussian process hyperparameter The likelihood function under the condition is Represents the hyperparameters of the Gaussian process The prior distribution of Among them, the sampling from the posterior distribution is done by the Metropolis-Hastings algorithm, and the sampling acceptance rate Defined as: ； in, represents the candidate new state generated by the current step, represents the improved Markov chain Monte Carlo method to optimize the hyperparameters of the Gaussian process model, Represents the hyperparameters after a given optimization The likelihood function in the case of Represents the hyperparameters of the Gaussian process The prior distribution of Select the next evaluation point using the improved Thompson sampling method , the sampling formula is: ； in, is the probability distribution of Thompson sampling, For the current observation data set, the expected improvement of the acquisition function is expressed as: ； in, Indicated in The objective function value at is the point with the best performance in the current observation set; The formula for updating the observation set is as follows: ； in, represents the observation noise; Retrain the Gaussian process model until the convergence condition is met and the optimal weight value of the adaptive enhancement model is obtained. and , and train the optimal classifier at the same time. The formula is as follows: ； in, Represents the final optimal classifier based on the integration of a series of weighted weak classifiers. To adjust the optimal parameters for model performance, represents a weighted linear combination of multiple weak classifiers, Indicates The weights of weak classifiers, Indicates A weak classifier.

2. The dual parameter adjustment adaptive enhancement method for cardiovascular disease diagnosis according to claim 1, characterized in that: The method also includes: preprocessing the acquired training data, including processing missing values, detecting outliers, data normalization and feature selection.

3. A dual-parameter adjustment adaptive enhancement device for cardiovascular disease diagnosis, characterized in that: include: A diagnosis module, used to input the acquired diagnosis sample data into the optimal classifier of the pre-acquired adaptive enhancement model to obtain diagnosis reference information; The method for obtaining the optimal classifier includes: Based on the acquired training data, train a weak classifier; Based on the trained weak classifier, an adaptive boosting model with adjustable parameter pairs is constructed; The adaptive boosting model is iteratively optimized using an improved Bayesian optimization method for parameter optimization to obtain the optimal classifier of the adaptive boosting model. The step of training a weak classifier based on the acquired training data includes: Input preprocessed training data : ； in, Representative The feature vector of the samples, represent Tags, ; represents the total number of samples; Initialize the weights of the training data using the following formula: ； ； in refers to the overall sample weight distribution, The first iteration The weights assigned to each sample; make ,in Indicates the sequence number of the current iteration. Refers to the total number of weak classifiers, and corresponds to the total number of iterations; The training dataset with weights is used to train weak classifiers , defined as follows: ； The weak classifier transforms the feature space The samples in the space are mapped to the label space ,in, Used to indicate whether a patient is diagnosed with cardiovascular disease. Indicates negative, Indicates positive; The method of constructing an adaptive enhancement model with adjustable parameter pairs based on the trained weak classifier includes: Compute the trained weak classifier The classification error rate on the training data set is as follows: ； In the formula, Indicates The feature vector of samples; Indicates The labels of samples, represents the classification error rate, It is The first iteration The weights assigned to each sample; is an indicator function, defined as follows: ； According to the classification error rate, calculate the weak classifier Weight , the formula is as follows: ； in, is the overall adjustment parameter for the weight update strength, Fine-tune parameters for update intensity; According to the weak classifier weights, update the weight distribution of the training data set, The weight distribution of the training data set in the iteration and In the iteration The weight of the samples The formula is as follows: ； in It is In the iteration The weight of the samples; ； in is the normalization factor, defined as follows: ； The weighted majority voting method is used to combine multiple base models into an adaptive enhanced model with adjustable parameter pairs. The formula is as follows: ； ； In the formula, is a linear combination of weighted weak classifiers, An adaptive boosting model representing a pair of adjustable parameters; The method of iteratively optimizing the adaptive enhancement model using the improved Bayesian optimization method for parameter optimization to obtain the optimal classifier of the adaptive enhancement model includes: Initialize the parameter observation set using Latin Hypercube sampling , the formula is as follows: ； in is the current parameter combination, is independent and identically distributed Gaussian noise, It means that the objective function is The value of the parameter combination; Use Gaussian process to train the parameter observation set The observed data in is modeled to obtain a Gaussian process model; The mean of the Gaussian process model and covariance function The formula is as follows: ； in Represents the variance of the observation noise; it is the variance parameter of Gaussian noise and is used to represent the uncertainty in the observation data; Represents the identity matrix, which is used to regularize the kernel matrix together with the noise variance to ensure stable calculation when the matrix is ​​inverted. is the covariance matrix, represents the target value set of the observed data, is the covariance function in the Gaussian process, which is used to represent the observation point currently considered and observation points The correlation between ； in, is the observation point currently considered, represents the predicted target point, and are the sets of observation points and objective function values, respectively. is the covariance function in the Gaussian process, used to represent the observation points and the predicted target point The correlation between Optimizing Hyperparameters of Gaussian Process Models Using Improved Markov Chain Monte Carlo Method , the posterior distribution It is expressed as: ； Represents a given Gaussian process hyperparameter The likelihood function under the condition is Represents the hyperparameters of the Gaussian process The prior distribution of Among them, the sampling from the posterior distribution is done by the Metropolis-Hastings algorithm, and the sampling acceptance rate Defined as: ； in, represents the candidate new state generated by the current step, represents the improved Markov chain Monte Carlo method to optimize the hyperparameters of the Gaussian process model, Represents the hyperparameters after a given optimization The likelihood function in the case of Represents the hyperparameters of the Gaussian process The prior distribution of Select the next evaluation point using the improved Thompson sampling method , the sampling formula is: ； in, is the probability distribution of Thompson sampling, For the current observation data set, the expected improvement of the acquisition function is expressed as: ； in, Indicated in The objective function value at is the point with the best performance in the current observation set; The formula for updating the observation set is as follows: ； in, represents the observation noise; Retrain the Gaussian process model until the convergence condition is met and the optimal weight value of the adaptive enhancement model is obtained. and , and train the optimal classifier at the same time. The formula is as follows: ； in, Represents the final optimal classifier based on the integration of a series of weighted weak classifiers. To adjust the optimal parameters for model performance, represents a weighted linear combination of multiple weak classifiers, Indicates The weights of weak classifiers, Indicates A weak classifier.

4. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method described in any one of claims 1 to 2 are implemented.

5. A computer device, characterized in that: include: Memory, for storing computer programs / instructions; A processor, configured to execute the computer program / instructions to implement the steps of the method according to any one of claims 1 to 2.

6. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 2 are implemented.

Images