Infectious disease patient sequelae risk prediction method based on machine learning

Abstract

The invention relates to an infectious disease patient sequelae risk prediction method based on machine learning, and the method comprises the following steps: obtaining multi-modal data of a patient, the multi-modal data comprising image data, text data and structured data; performing feature extraction on the multi-modal data to obtain related features; inputting the related features into a pre-constructed risk prediction model; performing sequelae risk prediction through the risk prediction model, and outputting a prediction result; and carrying out visual display on the prediction result. Risk prediction is carried out on the related features extracted based on the multi-modal data of the user through the risk prediction model constructed based on machine learning, an accurate prediction result can be obtained, and prediction support in the aspect is provided for related medical staff.

Description

Technical Field

[0001] The present invention relates to the technical field of machine learning, and particularly to a method for predicting the risk of sequelae in infectious disease patients based on machine learning. Background Art

[0002] During the COVID-19 pandemic, artificial intelligence technology has been widely used to analyze relevant data, and machine learning methods have played an important role in formulating diagnostic strategies, predicting epidemiological behaviors, and supporting the formulation and monitoring of public health policies. With the emergence of long COVID, machine learning methods have been used to develop prediction tools and build models for correlating patient phenotypes. The research directions at home and abroad mainly focus on using big data and machine learning methods to improve diagnostic accuracy and predict disease development. In terms of improving diagnosis, some studies have developed machine learning-based diagnostic models that combine patients' clinical manifestations, imaging data, and laboratory test results, aiming to improve the accuracy and efficiency of long COVID diagnosis. In terms of predicting disease development, the research mainly explores the application of machine learning algorithms in predicting the risk of long COVID. For example, the XGBoost strategy is often used to develop early prediction models for long COVID. These models can help doctors formulate early intervention plans to improve the curative effect and quality of life of patients.

[0003] Currently, there is a lack of a method for predicting the risk of sequelae in infectious disease patients based on machine learning, especially a method for accurately predicting the risk of different sequelae symptom groups, which is difficult to provide prediction support for early individualized intervention. Summary of the Invention

[0004] The purpose of the present invention is to at least solve one of the deficiencies of the prior art and provide a method for predicting the risk of sequelae in infectious disease patients based on machine learning.

[0005] To achieve the above purpose, the present invention adopts the following technical solutions:

[0006] Specifically, a method for predicting the risk of sequelae in infectious disease patients based on machine learning is proposed, including the following:

[0007] Obtain multimodal data of the patient, where the multimodal data includes image data, text data, and structured data;

[0008] Extract features from the multimodal data to obtain relevant features;

[0009] Input the relevant features into a pre-constructed risk prediction model;

[0010] Perform sequelae risk prediction through the risk prediction model and output a prediction result;

[0011] Visually display the prediction result.

[0012] Further, specifically, extracting features from the multi-modal data to obtain relevant features, including:

[0013] Extracting features from the image data to obtain the first data feature;

[0014] Extracting features from the text data to obtain the second data feature;

[0015] Extracting features from the structured data to obtain the third data feature.

[0016] Further, specifically, extracting features from the image data to obtain the first data feature, including:

[0017] Normalizing the image data to adjust the pixel values to the standard range, specifically as follows:

[0018]

[0019] where I is the original image data, I min and I max are the minimum and maximum values of the original image data respectively;

[0020] Performing random angle adjustment on I ′ i.e., I rot = rotate(I norm , θ),

[0021] Performing horizontal flipping on I rot i.e., processing I rot through the flip() function to obtain I rot ′ ,

[0022] Performing random cropping on I rot ′ to obtain I crop = crop(I rot ′ , crop_size), where crop_size is the preset cropping size,

[0023] Extracting features from I crop using the EfficientNet model pre-trained by transfer learning to obtain the high-dimensional image feature F img ,

[0024] F img = F EffNet (I crop ),

[0025] Performing on F imgPerform principal component analysis (PCA) processing to obtain F img-reduced ,

[0026] F img-reduced = PCA(F img ),

[0027] Reduce the dimension of F by finding the eigenvectors of the characteristic covariance matrix img-reduced ,

[0028] F img-reduced ′ = W T F img-reduced ,

[0029] where W T is the eigenvector matrix of, then F img-reduced ′ is the first data feature

[0030] Furthermore, specifically, extract data features for text data to obtain the second data feature, including

[0031] Preprocess the text data to remove stop words and punctuation marks to obtain the preprocessed text data Tokens

[0032] Then use the BioBERT model to perform deep semantic encoding on the preprocessed text data Tokens to extract the deep semantic feature F txt ,

[0033] F txt = F BioBERT (Tokens),

[0034] Perform average pooling operation on F txt to obtain the second data feature F txt-fixed ,

[0035]

[0036] Furthermore, specifically, extract data features for structured data to obtain the third data feature, including

[0037] Normalize the structured data to obtain X std ,

[0038] X st d = (X - μ X ) / σ X ,

[0039] where X represents the original structured data, μ X represents the mean of the original structured data, σ XRepresents the standard deviation of the original structured data;

[0040] Perform feature extraction on X through machine learning std to obtain feature F ml ,

[0041] F ml = MLFeatureExtraction(X std ),

[0042] Then, input feature F ml into the random forest model to extract the third data feature F str-selected ,

[0043]

[0044] Furthermore, specifically, the construction process of the risk prediction model includes

[0045] Fuse the first data feature, the second data feature, and the third data feature into a feature matrix to obtain F concat ,

[0046] F concat = [F img-reduced ′ , F txt-fixed , F str-selected ;

[0047] Perform multimodal fusion based on the multi-head attention mechanism. Specifically, superimpose F concat with the self-attention output through residual connection

[0048] H res = F concat + Z,

[0049] Then, perform normalization processing on the feature H res after residual connection through layer normalization to obtain the model output H out ,

[0050] H out = LayerNorm(H res ),

[0051] where Z = AV, A is the attention weight, and V is the weighted value vector

[0052]

[0053] Q = W Q F concat 、K = W K F concat 、V = W V Fconcat , W Q , W K , W V are preset weights respectively, k takes Q, K, V, d k represents the dimension of each head respectively,

[0054] Q, K, V satisfy

[0055] MultiHead(Q, K, V) = [Z1, Z2, …, Z h W O ,

[0056] Z i represents the output of the i-th attention head, i ∈ [1, h], h is the number of attention heads, W O is a linear transformation matrix.

[0057] Furthermore, the method further includes that the training process of the risk prediction model is as follows

[0058] The total loss function L includes cross-entropy loss and regression loss as follows

[0059] L = αL1 + βL2

[0060] where L1 and L2 represent cross-entropy loss and regression loss respectively

[0061] Use the Adam optimizer for training. The Adam optimizer adjusts parameter updates through an adaptive learning rate. The parameter update formula is as follows

[0062]

[0063] where θ, η, ∈ are Adam optimizer parameters, m t and v t are the first-order moment estimate and second-order moment estimate of the gradient respectively. The update rule is as follows

[0064] m t = β1m t-1 + (1 - β1)g t ,

[0065]

[0066] where g t is the gradient, β1 and β1 are momentum terms, and η is the learning rate.

[0067] Furthermore, the method further includes that when training the risk prediction model

[0068] introduce the feature-based gradient boosting machine F GBM, assuming the training set is X, the target variable is F, and the prediction of the model is:

[0069]

[0070] where h m (X) is the m-th decision tree, and α m is the weight of the m-th decision tree, M is the total number of decision trees,

[0071] Finally, the model is evaluated and corrected by Accuracy, Recall, Precision, and F1 Score.

[0072] Furthermore, the method further includes verifying the output prediction results through a questionnaire survey.

[0073] The beneficial effects of the present invention are:

[0074] The present invention proposes a method for predicting the sequela risk of infectious disease patients based on machine learning. By using a risk prediction model constructed based on machine learning to predict the risks of relevant features extracted from multi-modal data of users, relatively accurate prediction results can be obtained, providing prediction support in this regard for relevant medical staff. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] By describing the embodiments shown in the accompanying drawings in detail, the above and other features of the present disclosure will become more apparent. The same reference numerals in the drawings of the present disclosure denote the same or similar elements. Obviously, the drawings in the following description are only some embodiments of the present disclosure. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts. In the drawings:

[0076] Figure 1 Shows a flowchart of the method for predicting the sequela risk of infectious disease patients based on machine learning of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0077] The following will clearly and completely describe the concept, specific structure, and technical effects generated by the present invention in combination with the embodiments and the drawings, so as to fully understand the purpose, solution, and effects of the present invention. It should be noted that, without conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The same reference numerals used throughout the drawings indicate the same or similar parts.

[0078] Example 1, referring to Figure 1 , the present invention proposes a method for predicting the sequela risk of infectious disease patients based on machine learning, including the following:

[0079] Step 110: Obtain the multi-modal data of the patient, where the multi-modal data includes image data, text data, and structured data;

[0080] Step 120: Extract features from the multi-modal data to obtain relevant features;

[0081] Step 130: Input the relevant features into a pre-constructed risk prediction model;

[0082] Step 140: Perform sequela risk prediction through the risk prediction model and output the prediction result;

[0083] Step 150: Visualize and display the prediction result.

[0084] In this Embodiment 1, by using a risk prediction model constructed based on machine learning to perform risk prediction on the relevant features extracted from the user's multi-modal data, a relatively accurate prediction result can be obtained, providing prediction support in this aspect for relevant medical staff.

[0085] As a preferred embodiment of the present invention, specifically, extracting features from the multi-modal data to obtain relevant features includes:

[0086] Extract first data features from the image data;

[0087] Extract second data features from the text data;

[0088] Extract third data features from the structured data.

[0089] As a preferred embodiment of the present invention, specifically, extracting first data features from the image data includes:

[0090] Normalize the image data to adjust the pixel values to the standard range, specifically as follows:

[0091]

[0092] where I is the original image data, I min and I max are the minimum and maximum values of the original image data respectively; the purpose of this operation is to provide stable input data for subsequent analysis. In addition, by means of data augmentation technology, we have processed the images in a diversified manner, such as rotation, flipping, and cropping. These methods effectively expand the dataset and improve the model's performance on unknown data.

[0093] Perform random angle adjustment on I ′ i.e., I rot = rotate(Inorm , θ)

[0094] For I rot perform a horizontal flip, that is, process I through the flip() function rot to obtain I rot ′ ,

[0095] For I rot ′ perform random cropping to obtain I crop = crop(I rot ′ , crop_size), where crop_size is the preset cropping size

[0096] In order to capture deeper features from the preprocessed image, based on the EfficientNet model pre-trained by transfer learning, perform feature extraction on I crop to obtain high-dimensional image features F img ,

[0097] F img = F EffNet (I crop ),

[0098] Perform principal component analysis (PCA) on F img to obtain F img-reduced , aiming to reduce the feature dimension while retaining key image information, thereby reducing the computational cost and improving the training efficiency of the model;

[0099] F img-reduced = PCA(F img ),

[0100] Reduce the dimension of F by finding the eigenvectors of the feature covariance matrix img-reduced

[0101] F img-reduced ′ = W T F img-reduced ,

[0102] where W T is the eigenvector matrix, then F img-reduced ′ is the first data feature. By extracting the eigenvectors of the feature matrix, the high-dimensional data is compressed into a lower-dimensional representation form.

[0103] As a preferred embodiment of the present invention, specifically, perform data feature extraction on text data to obtain the second data feature, including

[0104] ​The feature extraction of text data covers medical records (such as patients' symptoms, medical history, and previous medications). In the preprocessing stage, we first removed stop words and punctuation marks. Then, the BioBERT model was used to perform deep semantic encoding on the processed text: BioBERT has been pre-trained on specialized biomedical corpora and can accurately capture the complex semantic structures and potential semantic associations in medical texts.

[0105] Preprocess the text data to remove stop words and punctuation marks to obtain preprocessed text data Tokens;

[0106] To further unify the dimensions of text features for fusion with other modal data, we introduced an average pooling operation to convert the extracted semantic information into a fixed-length vector representation:

[0107] Then, the BioBERT model was used to perform deep semantic encoding on the preprocessed text data Tokens to extract deep semantic features F txt ,

[0108] F txt =F BioBERT (Tokens),

[0109] For F txt Perform an average pooling operation to obtain the second data feature F txt-fixed ,

[0110]

[0111] As a preferred embodiment of the present invention, specifically, data feature extraction is performed on structured data to obtain a third data feature, including,

[0112] When processing structured data and feature extraction, we mainly deal with laboratory test results (such as various blood routine indicators, various biochemical indicators, inflammatory factor levels, etc.). To ensure that data with different dimensions can be compared on the same scale, we standardized these data: Standardize the structured data to obtain X std ,

[0113] X st d =(X - μ X ) / σ X ,

[0114] where X represents the original structured data, μ X represents the mean of the original structured data, and σ X represents the standard deviation of the original structured data; this standardization process effectively eliminates the dimensional differences between different features, enabling all data to be compared and modeled on a consistent basis.

[0115] For the standardized data, we adopted a variety of advanced machine learning methods for feature extraction. Through algorithms such as the random forest model and support vector machine, we can identify the most important features:

[0116] Feature extraction of X through machine learning std to obtain feature F ml ,

[0117] F ml = MLFeatureExtraction(X std ),

[0118] These features are further input into the random forest model to extract and integrate the important features of the structured data for output:

[0119] Then, input feature F ml into the random forest model to extract the third data feature F str-selected ,

[0120]

[0121] where represents the random forest model constructed based on the tree parameter k, and its input is feature F ml As a preferred embodiment of the present invention, specifically, the construction process of the risk prediction model includes,

[0122] Multimodal Fusion is to integrate data from different modalities (such as images, texts, structured data, time series data, etc.) to make full use of the complementarity of multi-source information and improve the overall performance of the task. By fusing data of multiple modalities, the system can capture more context and features, improve the accuracy and robustness of tasks such as classification and detection, and thus provide more reliable decision support in complex scenarios. To achieve the effective fusion of multimodal data, we adopted the self-attention mechanism (Transformer). First, integrate the image features, text features, and structured data features, as well as the correlation features extracted by the prediction model and the large language model, into a feature matrix: fuse the first data feature, the second data feature, and the third data feature into a feature matrix to obtain F concat ,

[0123] F concat = [F img-reduced ′ , F txt-fixed , F str-selected ;

[0124] The process of feature fusion relies on the self-attention mechanism. This mechanism determines which modal features are most important for the current task by calculating the correlation weights between each feature and other features. The specific operations are as follows: Multimodal fusion is performed based on the multi-head attention mechanism to obtain fused data. Specifically, F concat is superimposed with the self-attention output,

[0125] H res = F concat + Z,

[0126] Then, the feature H res after the residual connection is normalized through layer normalization to obtain the model output H out ,

[0127] H out = LayerNorm(H res ),

[0128] where Z = AV, A is the attention weight, and V is the weighted value vector,

[0129]

[0130] Q = W Q F concat 、K = W K F concat 、V = W V F concat , W Q 、W K 、W V are preset weights respectively, k takes Q, K, V, and d k represents the dimension of each head respectively,

[0131] Q, K, V satisfy,

[0132] MultiHead(Q,K,V) = [Z1,Z2,…,Z h W O ,

[0133] Z i represents the output of the i-th attention head, i ∈ [1, h], h is the number of attention heads, and W O is the linear transformation matrix.

[0134] As a preferred embodiment of the present invention, the method further includes the following training process for the risk prediction model,

[0135] To achieve accurate classification and prediction of depression or fatigue sequelae, we adopted a multi-task learning framework to output two parts: etiological diagnosis and symptom categories. The total loss function of multi-task learning includes cross-entropy loss and regression loss: The total loss function L includes cross-entropy loss and regression loss as follows,

[0136] L = αL1 + βL2

[0137] where L1 and L2 represent cross-entropy loss and regression loss respectively,

[0138] Training is performed using the Adam optimizer, which adjusts parameter updates through an adaptive learning rate. The parameter update formula is as follows:

[0139]

[0140] where θ, η, ∈ are Adam optimizer parameters, m t and v t are the first and second moment estimates of the gradient respectively, and the update rules are as follows:

[0141] m t = β1m t-1 + (1 - β1)g t 、

[0142]

[0143] where g t is the gradient, β1 and β1 are momentum terms, and η is the learning rate.

[0144] As a preferred embodiment of the present invention, the method further includes, when training the risk prediction model,

[0145] In addition, in addition to the core model, we also introduced a feature-based gradient boosting machine (F GBM ) to further enhance prediction accuracy and robustness. F GBM combines traditional gradient boosting and feature engineering and is particularly suitable for processing complex multi-modal data. The GBM model is trained by minimizing a loss function (such as logarithmic loss or mean squared error). Introducing the feature-based gradient boosting machine F GBM , assuming the training set is X, the target variable is F, and the prediction of the model is:

[0146]

[0147] where h m (X) is the m-th decision tree, α m is the weight of the m-th decision tree, and M is the total number of decision trees,

[0148] FGBM By combining a decision tree (as a weak learner) with the features extracted by the self-attention layer, it is possible to capture the non-linear relationships and interactions between features, which may not be fully expressed by linear models. FGBM makes the final prediction by combining the output results of multiple decision trees, reducing the bias and variance of the model, thereby improving the accuracy and stability of the prediction.

[0149] During the training process, the FGBM model updates its parameters by minimizing the loss function, usually in combination with gradient descent and regularization techniques to prevent overfitting. The decision trees in FGBM can capture the complex interactions between different modalities, improving the prediction ability of the model in practical applications.

[0150] Validation of the Risk Model

[0151] When evaluating a risk model, the key validation metrics include Accuracy, Precision, Recall, and F1-Score.

[0152] Accuracy: Represents the proportion of samples correctly classified by the model to the total samples. Calculated according to the following formula:

[0153]

[0154] Where TP is the true positive sample size, TN is the true negative sample size, FP is the false positive sample size, and FN is the false negative sample size.

[0155] Precision: Measures how many of the samples predicted as positive are truly positive. Formula:

[0156]

[0157] Recall: Measures how many of the positive class samples are predicted as positive. Formula:

[0158]

[0159] F1-Score: Considers both Precision and Recall. Formula:

[0160]

[0161] Through the comprehensive evaluation of the above metrics, the effectiveness and reliability of the prediction model can be fully understood.

[0162] Finally, the model is evaluated and corrected using Accuracy, Recall, Precision, and F1 Score.

[0163] As a preferred embodiment of the present invention, the method further includes authenticating the output prediction results through a questionnaire survey.

[0164] Specifically, in application,

[0165] Assessment of sequelae

[0166] Subjects will be required to complete a comprehensive neuropsychological assessment, including the Patient Health Questionnaire (PHQ-9) (Table 1) and the Fatigue Assessment Scale (FAS) (Table 2). The FAS assesses a total of 10 items, of which 5 reflect physical fatigue and 5 reflect mental fatigue.

[0167] Table 1. Patient Health Questionnaire

[0168]

[0169] Table 2. Fatigue Assessment Scale

[0170]

[0171]

[0172] Through the above Table 1 and Table 2, the outcome indicators are the presence or absence of depression or fatigue sequelae at 12 months. As mentioned above, the PHQ-9 scale is used to assess depression, and the main statistical indicator is the total score. 0-4 points indicate no depressive symptoms, 5-9 points indicate mild depressive symptoms, 10-14 points indicate moderate depressive symptoms, and 15 points and above indicate severe depressive symptoms. As mentioned above, the FAS scale is used to assess fatigue. Below 22 points indicates no fatigue symptoms, and 22 points and above indicate obvious fatigue symptoms.

[0173] Although the description of the present invention has been quite detailed and particularly describes several of the described embodiments, it is not intended to be limited to any of these details or embodiments or any particular embodiment, but should be regarded as providing a broad interpretation of these claims in light of the prior art by reference to the appended claims, thereby effectively covering the intended scope of the present invention. In addition, the present invention has been described above in terms of embodiments foreseeable by the inventor for the purpose of providing a useful description, and those non-substantive modifications to the present invention that are not currently foreseeable may still represent equivalent modifications of the present invention.

[0174] As mentioned above, these are only the preferred embodiments of the present invention. The present invention is not limited to the above-mentioned embodiments. As long as the same means are used to achieve the technical effects of the present invention, they should fall within the protection scope of the present invention. Within the protection scope of the present invention, various modifications and variations can be made to its technical solutions and / or embodiments.

Claims

1. A method for predicting the risk of sequelae in infectious disease patients based on machine learning, characterized in that, Including the following: Obtain multimodal data of a patient, where the multimodal data includes image data, text data, and structured data; Extract features from the multimodal data to obtain relevant features; Input the relevant features into a pre-constructed risk prediction model; Perform sequela risk prediction through the risk prediction model and output a prediction result; Visually display the prediction result.

2. The method for predicting the sequela risk of infectious disease patients based on machine learning according to claim 1, wherein Specifically, extracting features from the multimodal data to obtain relevant features includes: Extract first data features from the image data; Extract second data features from the text data; Extract third data features from the structured data.

3. The method for predicting the sequela risk of infectious disease patients based on machine learning according to claim 2, wherein Specifically, extracting first data features from the image data includes: Normalize the image data to adjust the pixel values to the standard range, as shown in the following formula: where I is the original image data, I min and I max are the minimum and maximum values of the original image data, respectively; Randomly adjust the angle of I′, that is, I rot = rotate(I norm , θ) For I rot perform a horizontal flip, that is, process I rot through the flip() function to obtain I rot ′. For I rot ′, perform random cropping to obtain I crop = crop(I rot ′, crop_size), where crop_size is the preset cropping size Based on the EfficientNet model pre-trained by means of transfer learning, perform feature extraction on I crop to obtain high-dimensional image features F img , F img = F EffNet (I crop ) For F img Perform principal component analysis (PCA) processing on it to obtain F img-reduced , F img-reduced = PCA(F img ) Reduce the dimension of F by finding the eigenvectors of the characteristic covariance matrix img-reduced dimensionality reduction F img-reduced ′ = W T F img-reduced , Among them, W T is the feature vector matrix, then F img-reduced ′ is the first data feature.

4. The method for predicting the sequela risk of infectious disease patients based on machine learning according to claim 3, wherein, Specifically, extracting second data features for the text data includes: Preprocess the text data to remove stop words and punctuation marks to obtain preprocessed text data Tokens; Next, the BioBERT model is used to perform deep semantic encoding on the preprocessed text data Tokens to extract deep semantic features F txt , F txt = F BioBERT (Tokens) For F txt Perform average pooling operation to obtain the second data feature F txt-fixed , 5. The method for predicting the risk of sequelae in infectious disease patients based on machine learning according to claim 4, characterized in that, Specifically, extracting third data features for the structured data includes: Standardize the structured data to obtain X std , X st d =(X - μ X ) / σ X , where X represents the original structured data, μ X represents the mean of the original structured data, and σ X represents the standard deviation of the original structured data; Feature extraction of X through machine learning std to identify the most important feature F ml , F ml = MLFeatureExtraction(X std ) Then input feature F ml into the random forest model to extract the third data feature F str-selected , 6. The method for predicting the sequela risk of infectious disease patients based on machine learning according to claim 5, wherein, Specifically, the construction process of the risk prediction model includes: Fuse the first data feature, the second data feature, and the third data feature into a feature matrix to obtain F concat , F concat = [F img-reduced ', F txt-fixed , F str-selected ; Multi-modal fusion is performed based on the multi-head attention mechanism to obtain fused data. Specifically, F is superimposed with the self-attention output through residual connection. concat ​ H res = F concat + Z, Then, the feature H after residual connection is normalized through layer normalization res to obtain the model output H out , H out = LayerNorm(H res ) Where Z = AV, A is the attention weight, and V is the weighted value vector, Q = W Q F concat 、K = W K F concat 、V = W V F concat ,W Q 、W K 、W V are preset weights respectively, k takes Q, K, V, d k represents the dimension of each head respectively Q, K, V satisfy: MultiHead(Q,K,V)=[Z1,Z2,…,Z h W O , Z i represents the output of the i-th attention head, where i ∈ [1, h] and h is the number of attention heads, and W O is a linear transformation matrix.

7. The method for predicting the risk of sequelae in infectious disease patients based on machine learning according to claim 6, wherein The method further includes the following training process for the risk prediction model: The total loss function L includes cross-entropy loss and regression loss as follows: L = αL1 + βL2 Where L1 and L2 respectively represent cross-entropy loss and regression loss, Use the Adam optimizer for training. The Adam optimizer adjusts parameter updates through an adaptive learning rate, and the parameter update formula is as follows: where θ, η, ∈ are Adam optimizer parameters, and m t and v t are the first and second moment estimates of the gradient respectively, and the update rules are as follows: m t = β1m t-1 + (1 - β1)g t , where g t is the gradient, β1 and β1 are momentum terms, and η is the learning rate.

8. The method for predicting the risk of sequelae in infectious disease patients based on machine learning according to claim 7, wherein The method further includes, when training the risk prediction model, Introduce the feature-based gradient boosting machine F GBM , assuming the training set is X, the target variable is F, and the prediction of the model is: where h m (X) is the m-th decision tree, and α m is the weight of the m-th decision tree, and M is the total number of decision trees, Finally, evaluate and correct the model through accuracy Accuracy, recall Recall, precision Precision, and F1 score F1 Score.

9. The method for predicting the risk of sequelae in infectious disease patients based on machine learning according to claim 8, wherein The method further includes, for the output prediction result, conduct result certification through a questionnaire survey.

Images