Disease prediction method and system based on causal inference and dynamic integrated multi-label

Abstract

The application provides a disease prediction method and system based on causal inference and dynamic integrated multi-label, the method comprises the following steps: acquiring multi-source information of a patient, including demographic indicators, lifestyle, physical examination, chief complaint, medical history and previous medication information; establishing a causal model to analyze the causal relationship between each feature and selecting a feature set with causal effect; training multiple multi-label base learning classifiers using the feature set with causal effect, updating the weights through stacked integration to obtain the best-performing prediction model; through different numbers, types and combination methods of the best-performing prediction model, a new multi-label integrated prediction model is dynamically constructed, and the combination model with the highest prediction performance is selected to predict the disease. The prediction method provided by the application can reflect the causal relationship between each feature, avoid misjudgment caused by correlation, improve prediction accuracy, help improve the diagnosis level and treatment effect of doctors, and promote the digitalization and intelligentization development of the medical industry.

Description

Technical Field

[0001] This invention relates to the field of disease prediction technology, specifically to a disease prediction method and system based on causal inference and dynamic integration of multi-labels. Background Technology

[0002] Currently, the prediction of various chronic diseases mainly relies on traditional medical and biostatistical methods. These methods are typically based on sample datasets and utilize machine learning or artificial intelligence algorithms, such as support vector machines, decision trees, and neural networks, to predict the risk of disease onset. However, these methods have some limitations, such as their inability to consider the interaction of multiple factors and their evolution over time; therefore, the accuracy and reliability of the predictions are somewhat limited.

[0003] Traditional machine learning algorithms often overlook underlying causal relationships when predicting diseases, which can lead to biased predictions. Furthermore, disease progression is typically dynamic, and the time factor plays a crucial role in disease prediction. Currently, disease prediction plays a vital role in clinical practice. Traditional disease prediction methods often focus only on specific disease indicators or symptoms, neglecting the complex relationships between different factors. Therefore, misdiagnosis and missed diagnosis frequently occur during diagnosis and treatment. Summary of the Invention

[0004] Therefore, the technical problem to be solved by the present invention is to overcome the defects existing in the prior art, thereby providing a disease prediction method and system based on causal inference and dynamic integration of multi-labels. By using causal inference algorithms to analyze the causal relationships between different data, the accuracy of disease prediction is improved; at the same time, the use of dynamic integration of multi-label algorithms for training and prediction improves the robustness and generalization performance of the algorithm.

[0005] The technical solution of the present invention to solve the above-mentioned technical problems is as follows:

[0006] In a first aspect, the present invention provides a disease prediction method based on causal inference and dynamic integration of multi-labels, comprising the following steps:

[0007] Obtain multi-source information from patients, including: demographic information, lifestyle information, physical examination information, chief complaints, past medical history and past medication information;

[0008] Establish a causal model to analyze the causal relationships between various features and screen out the feature set with causal effects;

[0009] Multiple multi-label base learning classifiers are trained using the feature set with causal effects, and the weights are updated through stacked ensemble to obtain the best-performing prediction model.

[0010] By combining different numbers and types of optimal predictive models, new multi-label integrated predictive models are dynamically constructed, and the combined model with the highest predictive performance is selected to predict diseases.

[0011] Optionally, the demographic information may include: gender, age, height, and weight; lifestyle information may include: smoking history and drinking history; physical examination information may include: biochemical tests, electrocardiogram information, and imaging data; and past medical history information may include: the patient's personal medical history and family medical history.

[0012] Optionally, the step of establishing a causal model to analyze the causal relationships between various features and screening out a feature set with causal effects includes:

[0013] Constructing a Bayesian network: Let probability P(U) be the joint probability distribution of outcome y, y∈L, n=1,...,N, where N represents the number of patients, L={l1,l2,...,lq} be a set of q distinct binary outcome labels, and U be the set of nodes G in a directed acyclic graph. If G...<U,G,P(U)> If the Markov condition is satisfied, then<U,G,P(U)> The triples are called Bayesian networks, where each variable is independent of any subset of non-child terms in G, given the parent terms.

[0014] Training Markov Chains: Setting BN<U,G,P(U)> In a Bayesian network based on the loyalty assumption, F i A Markov blanket ∈ F is denoted as MB(Fi), where MB(F) i )={pa(F i )Uch(F i )Usp(F i )} is a unique term, F represents different features, pa(F i ) represents F i The set of parent nodes that directly affect F i The set of variables; ch(F i ) represents F i The set of child nodes, i.e., F i The set of variables directly affected; sp(F i ) indicates that it is related to F i The set of other nodes that share the same parent node, and F i A set of variables that have an indirect influence relationship;

[0015] Filtering multi-label association features: For feature set F, the outcome probability P(T) i Maximize |S), where The causal feature selection for dataset D is defined as follows:

[0016] S * =arrgmax|S|,

[0017] stP i (T i S)=P′(T i S)(T i GT′ i ,,j≠i)

[0018] Where T represents the possible disease category to be output;

[0019] Repeat the above process of training the Markov chain and filtering multi-label associated features, and finally maximize the probability distribution of the most feature corresponding to all outcome labels y∈L, and select the feature set with causal effect.

[0020] Optionally, the process of training multiple multi-label base learning classifiers includes:

[0021] Initialization: For all patient individuals i, initialize the weight W1(i,l) for each patient, obtain the initial sample dataset D1, where l represents the label, and set the iteration number t=1;

[0022] Training base classifiers: Use the m-th base classifier to perform stacked ensemble on dataset D1 to train a single base classifier h. m1 (x,l) predicts patient outcomes;

[0023] Weight update: Calculate the Hamming loss of the base classifier, i.e., the proportion of misclassified labels e. t Calculate and update weight α t Using α t Calculate the next iteration update W t+1 (i,l);

[0024] Repeated integration iteration: Set the number of iterations t = t + 1, until the preset number of iterations is reached;

[0025] Weighted voting for learning classifiers: selecting a single classifier h for t = 1, ..., T. mt Weighted voting is used to improve the learning classifier h. m .

[0026] Optionally, the step of dynamically constructing a new multi-label ensemble prediction model by combining different numbers and types of base classifiers includes:

[0027] Initialization: The original dataset D with causal effects s It is an empty set;

[0028] Classify the training samples: Use the trained base learning classifier h m (x,l) classifies feature x1 to obtain c 1m =hm (x 1, l);

[0029] Update dataset D s :D s ={c 11 ,c 12 ,…,c 1m The process of classifying the training samples is repeated until classification predictions are performed for all N hospitalized patients, and c is obtained. nm =h m (x n ,l), to obtain a new dataset D s ={((c i1 ,c i2 ,…,c im ),y)};

[0030] Ensemble part of training the model: using the new dataset D s The training model results are dynamically selected based on the base learner pool, and finally stacked and ensembled. This part uses a new learning algorithm Z, then H = Z(D s );

[0031] Output: H(x) = H(h1(x,l),h2(x,l),…,h m (x,l)).

[0032] Optionally, after obtaining the patient's multi-source information, the process also includes:

[0033] The multi-source information is preprocessed and cleaned to remove noise and outliers, and feature selection and dimensionality reduction are performed.

[0034] Secondly, embodiments of the present invention provide a disease prediction system based on causal inference and dynamic integration of multiple labels, the system comprising:

[0035] The data collection module is used to acquire multi-source information about patients, including: demographic information, lifestyle information, physical examination information, chief complaints, past medical history and past medication information;

[0036] The causal inference module is used to analyze the causal relationships between various features to build a causal model, and to select a feature set with causal effects based on the causal inference model.

[0037] The dynamic ensemble multi-label algorithm module is used to train multiple multi-label base learning classifiers using the feature set with causal effects, and update the weights through stacked ensemble to obtain the best-performing prediction model.

[0038] The disease prediction model determination module is used to dynamically construct new multi-label integrated prediction models by combining different numbers and types of optimal prediction models, and select the combination model with the highest prediction performance to predict diseases.

[0039] Thirdly, embodiments of the present invention provide a computer device, including: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the method described in the first aspect, or any optional embodiment of the first aspect.

[0040] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing computer instructions for causing the computer to perform the method described in the first aspect, or any optional embodiment of the first aspect.

[0041] This invention provides a disease prediction method and system based on causal inference and dynamic integration of multi-labels. By analyzing the causal relationships and dynamic changes among different patient characteristics and combining them with a multi-label classification algorithm, it can reflect the causal relationships between various characteristics, avoid misjudgments caused by correlation, and improve the accuracy of disease prediction. This can help improve doctors' diagnostic skills and treatment outcomes, and contribute to promoting the digital and intelligent development of the medical industry. Furthermore, the disease prediction method and system provided by this invention have broad application prospects and are of great significance in fields such as healthcare, health management, and medical research. Attached Figure Description

[0042] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0043] Figure 1 A flowchart illustrating a disease prediction method based on causal inference and dynamic integration of multiple labels provided in an embodiment of the present invention;

[0044] Figure 2 A schematic diagram illustrating the principle of a disease prediction method based on causal inference and dynamic integration of multiple labels, provided in an embodiment of the present invention.

[0045] Figure 3 A flowchart illustrating the key steps of a disease prediction method based on causal inference and dynamic integration of multiple labels, provided in an embodiment of the present invention;

[0046] Figure 4 A schematic diagram of the module composition of a disease prediction system based on causal inference and dynamic integration of multiple labels provided in an embodiment of the present invention;

[0047] Figure 5 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0049] Furthermore, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0050] This invention provides a disease prediction method based on causal inference and dynamic integration of multi-labels, the flowchart of which is shown below. Figure 1 As shown, the principle block diagram is as follows: Figure 2 As shown:

[0051] Step S1: Obtain multi-source information about the patient, including: demographic information, lifestyle information, physical examination information, chief complaint, past medical history and past medication information.

[0052] Specifically, in this embodiment of the invention, demographic information includes: gender, age, height, weight, and occupation; lifestyle information includes: smoking history and alcohol consumption history; physical examination information includes: biochemical tests (white blood cells, red blood cells, red blood cell distribution width, hemoglobin, platelets, etc.), electrocardiogram information, and imaging data; past medical history information includes: the patient's personal medical history and family medical history. The chief complaint is a clinical manifestation actively described by the patient, such as chest pain, nausea, and palpitations.

[0053] Step S2: Establish a causal model to analyze the causal relationships between various features and select the feature set with causal effects.

[0054] Specifically, in this embodiment of the invention, the collected patient information is used to construct a multi-label causal feature selection framework based on the principle of causal invariance. The concept of Markov blanket (MB) in Bayesian networks is used to give the final feature set of the multi-label data source with causal effect. The specific steps include:

[0055] Step S21: Construct a Bayesian network: Let probability P(U) be the joint probability distribution of outcome y, y∈L, n=1,...,N, where N represents the number of patients, L={l1,l2,...,lq} is a set of q distinct binary outcome labels, and U is the set of nodes G in a directed acyclic graph. If G...<U,G,P(U)> If the Markov condition is satisfied, then<U,G,P(U)> The triples are called Bayesian networks, where each variable is independent of any subset of non-child terms in G, given the parent terms.

[0056] Step S22: Training the Markov chain: Setting up BN<U,G,P(U)> For loyalty-based Bayesian networks, F i A Markov blanket ∈ F is denoted as MB(Fi), where MB(F) i )={pa(F i )Uch(F i )Usp(F i )} is a unique term, F represents different features, pa(F i ) represents F i The set of parent nodes that directly affect F i The set of variables; ch(F i ) represents F i The set of child nodes, i.e., F i The set of variables directly affected; sp(F i ) indicates that it is related to F i The set of other nodes that share the same parent node, and F i A set of variables that have an indirect influence relationship.

[0057] Step S23: Filter multi-label association features: For feature set F, calculate the outcome probability P(T) i Maximize |S), where The causal feature selection for dataset D is defined as follows:

[0058] S * =arrgmax|S|,

[0059] stP i (T i S)=P′(T i S)(T i GT i ′ , ,j≠i)

[0060] Where T represents the possible disease category to be output;

[0061] Step S24: Repeat steps S22-S23 above to maximize the probability of the most characteristic distribution corresponding to all outcome labels y∈L, and select the feature set with causal effect.

[0062] Steps S21-S22 construct a causal network graph, building causal chains from the patient's multi-source information and selecting features that may have a causal effect on disease prediction. Step S23 combines potential disease labels with the previously selected variables to further filter for features with a causal effect on the disease labels. This process effectively reduces the dimensionality of patient information, filters out features with causal effects for subsequent model construction, and ensures that the selected features have a genuine causal effect on the labels.

[0063] Step S3: Train multiple multi-label base learning classifiers using the feature set with causal effects, and update the weights through stacked ensemble to obtain the best-performing prediction model.

[0064] The goal of step S3 is to find a base multi-label classifier to improve the model's prediction performance. Since there are many types of multi-label algorithms, selecting a base classifier is challenging. In this embodiment, C4.5 is used as the meta-classifier, combined with four relatively stable multi-label models—BR, CC, LP, and RAKEL—to construct a base classifier. Weights are updated through stacking ensemble (hereinafter referred to as Stacking) to obtain the best-performing prediction model. In this embodiment, for m = 1, ..., M, training M multi-label base learning classifiers specifically includes the following steps:

[0065] Step S31: Initialization: For all patient individuals i, initialize the weight W1(i,l) = 1 / N for each patient, obtain the initial sample dataset D1, and set the iteration number t = 1;

[0066] Step S32: Train the base classifier: Use the m-th base classifier to perform stacked ensemble on dataset D1 to train a single base classifier h. m1 (x,l) predicts patient outcomes;

[0067] Step S33: Weight Update: Calculate the Hamming loss of the base classifier (the smaller the value, the better the model's prediction performance), i.e., the misclassified label ratio e. t Calculate and update weights Using α t Calculate the next iteration update As a normalization factor, y[l] indicates whether the label l belongs to instance (x,y).

[0068] Step S34: Repeat the integration iteration: Set the iteration number t = t + 1 until the preset iteration number T is reached;

[0069] Step S35: Weighted voting for the learned classifier: Select a single classifier h for t = 1, ..., T. mtWeighted voting is used to improve the learning classifier h. m , as the best performing prediction model.

[0070] Step S4: Dynamically construct new multi-label integrated prediction models by combining different numbers and types of optimal prediction models, and select the combination model with the highest prediction performance to predict the disease.

[0071] The process of dynamically constructing a new multi-label ensemble prediction model according to embodiments of the present invention includes the following steps:

[0072] Step S41: Initialization: Original dataset D with causal effects s It is an empty set;

[0073] Step S42: Classify the training samples: Use the trained base learning classifier h m (x,l) classifies feature x1 to obtain c 1m =h m (x1,l);

[0074] Step S43: Update dataset D s :D s ={c 11 ,c 12 ,…,c 1m The process of classifying the training samples is repeated until classification predictions are performed for all N hospitalized patients, and c is obtained. nm =h m (x n ,l), to obtain a new dataset D s ={((c i1 ,c i2 ,…,c im ),y)};

[0075] Step S44: Ensemble part of training the model: using the new dataset D s The training model results are dynamically selected based on the base learner pool, and finally stacking is performed. This part uses a new learning algorithm Z, then H = Z(D s );

[0076] Step S45: Output: H(x) = H(h1(x,l),h2(x,l),…,h m (x,l)).

[0077] The above process dynamically constructs new multi-label ensemble prediction models by combining different numbers and types of base classifiers, selects the best-performing base classifiers, and makes a final diagnosis of the patient's disease to ensure the highest accuracy of the predictions.

[0078] The flowchart of the key steps of the disease prediction method based on causal inference and dynamic integration of multi-labels provided in this embodiment of the invention is as follows: Figure 3 As shown, it has good versatility and adaptability, and can be applied to various disease prediction scenarios. It can also adaptively update the prediction model according to changes in data, maintaining the accuracy and efficiency of prediction.

[0079] This invention provides a disease prediction system based on causal inference and dynamic integration of multi-labels, such as... Figure 4 As shown, the system includes:

[0080] The data collection module is used to acquire multi-source information about the patient, including: demographic information, lifestyle information, physical examination information, chief complaints, past medical history, and past medication information. For details, please refer to the relevant description of step S1 in the above method embodiments, which will not be repeated here.

[0081] The data processing module is used to preprocess and clean the multi-source information, remove noise and outliers, perform feature selection and dimensionality reduction operations, and prepare for the subsequent construction of disease prediction models.

[0082] The causal inference module is used to analyze the causal relationships between various features to establish a causal model, and to select a feature set with causal effects based on the causal inference model. For details, please refer to the relevant description of step S2 in the above method embodiment, which will not be repeated here.

[0083] A dynamic multi-label ensemble module is used to train multiple multi-label base learning classifiers using the feature set with causal effects, and to update the weights through stacked ensemble to obtain the best-performing prediction model. For details, please refer to the relevant description of step S3 in the above method embodiment, which will not be repeated here.

[0084] The disease prediction model determination module is used to dynamically construct new multi-label integrated prediction models by combining different numbers and types of optimal prediction models, and select the combination model with the highest predictive performance to predict the disease. For details, please refer to the relevant description of step S4 in the above method embodiments, which will not be repeated here.

[0085] The interface display module is used to present the disease prediction results to users. Through an intuitive user interface, users can more easily obtain the prediction results and make corresponding clinical decisions.

[0086] In a preferred embodiment, the system provided by the present invention further includes a disease prediction model update module, comprising an incremental learning unit and a transfer learning unit. In the incremental learning unit, the weights of new samples and old samples are adjusted using a method based on sample importance. In the transfer learning unit, existing prediction models are used to predict new diseases. By learning and receiving new data, the module can continuously ensure the accuracy of disease prediction.

[0087] Examples of practical applications using the disease prediction system provided in this invention:

[0088] 1. Patient A, female, 45 years old, height 160cm, weight 70kg, no history of smoking, history of alcohol consumption, no family history of heart disease; chief complaint: recent onset of palpitations and shortness of breath; the patient had previously undergone an electrocardiogram, which revealed arrhythmia; biochemical test results are as follows: white blood cell count 8.5×10^9 / L, red blood cell count 4.1×10^12 / L, red blood cell distribution width 13.6%, hemoglobin 131g / L, platelet count 226×10^9 / L. Imaging data showed signs of left ventricular hypertrophy.

[0089] The doctor uses the system provided in this embodiment of the invention to input patient A's gender, age, height, weight, smoking history, drinking history, family history, clinical manifestations (symptoms such as palpitations and shortness of breath), biochemical tests (white blood cells, red blood cells, red blood cell distribution width, hemoglobin, platelets, etc.), electrocardiogram information, imaging data, and other information. The system filters out disease-related features, then trains and predicts the model, ultimately obtaining the following disease prediction results:

[0090] High blood pressure: The probability of developing the condition is 70%.

[0091] Cardiac arrhythmia: The probability of developing the condition is 60%.

[0092] Coronary heart disease: The probability of developing the disease is 40%.

[0093] Based on patient A's symptoms, signs, and examination results, the doctor further diagnosed the disease and ultimately determined it to be arrhythmia. A corresponding treatment plan was then developed, including medication and lifestyle adjustments.

[0094] 2. In intelligent medical devices, the device can use sensors to monitor the patient's physiological indicators (such as heart rate, blood pressure and blood oxygen saturation), and transmit these data to a cloud server that integrates the disease prediction system provided in this embodiment for analysis and diagnosis, and automatically send alarms or reminders to doctors when necessary based on the diagnosis results.

[0095] Figure 5A schematic diagram of a computer device according to an embodiment of the present invention is shown, including: a processor 901 and a memory 902, wherein the processor 901 and the memory 902 can be connected via a bus or other means. Figure 5 Taking the example of a connection between China and Israel via a bus.

[0096] Processor 901 can be a Central Processing Unit (CPU). Processor 901 can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or combinations of the above types of chips.

[0097] The memory 902, as a non-transitory computer-readable storage medium, can be used to store non-transitory server programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the methods in the above method embodiments. The processor 901 executes various functional applications and data processing of the processor by running the non-transitory server programs, instructions, and modules stored in the memory 902, thereby implementing the methods in the above method embodiments.

[0098] The memory 902 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created by the processor 901, etc. Furthermore, the memory 902 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, the memory 902 may optionally include memory remotely located relative to the processor 901, and these remote memories may be connected to the processor 901 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0099] One or more modules are stored in memory 902, and when executed by processor 901, they perform the methods described in the above method embodiments.

[0100] The specific details of the aforementioned computer equipment can be understood by referring to the relevant descriptions and effects in the above method embodiments, and will not be repeated here.

[0101] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The implemented program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk drive (HDD), or solid-state drive (SSD), etc.; the storage medium can also include combinations of the above types of memory.

[0102] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A disease prediction method based on causal inference and dynamic integration of multi-labels, characterized in that, Includes the following steps: Obtain multi-source information from patients, including: demographic information, lifestyle information, physical examination information, chief complaints, past medical history and past medication information; Establish a causal model to analyze the causal relationships between various features and screen out the feature set with causal effects; Multiple multi-label base learning classifiers are trained using the feature set with causal effects, and the weights are updated through stacked ensemble to obtain the best-performing prediction model. By dynamically constructing new multi-label integrated prediction models through different combinations of optimal-performing prediction models of varying numbers and types, and selecting the combination model with the highest predictive performance to predict diseases, the steps of establishing a causal model to analyze the causal relationships between various features and screening out the feature set with causal effects include: Building Bayesian Networks: Let Let L be the joint probability distribution of outcome y, y∈L, n = 1,..., N, where N represents the number of patients, L = {l1,l2,...,lq} is a set of q distinct binary outcome labels, and U is the set of nodes G in a directed acyclic graph. If G...<U, G, P(U)> If the Markov condition is satisfied, then<U, G, P(U)> The triples are called Bayesian networks, where each variable is independent of any subset of non-child terms in G, given the parent terms. Training Markov Chains: Setting BN<U, G, P(U)> For loyalty-based hypothesis Bayesian networks A Markov blanket ∈ F is denoted as MB(Fi), where For a unique term, F represents a distinct feature. express The set of parent nodes, i.e., those directly affected The set of variables; express The set of child nodes, i.e. The set of variables that are directly affected; Indicates and The set of other nodes that share the same parent node, and A set of variables that have an indirect influence relationship; Filtering multi-label association features: targeting feature sets probability of the outcome Maximize, where , The causal feature selection for dataset D is defined as follows: Where T represents the possible disease category to be output; Repeat the above process of training Markov chains and filtering multi-label association features, and finally maximize the probability distribution of the most feature corresponding to all outcome labels y∈L, and select the feature set with causal effect; The process of training multiple multi-label base learning classifiers includes: Initialization: For all individual patients i Initialize the weights for each patient. W 1( i, l ), obtain the initial sample dataset D 1, l Represents the label and sets the number of iterations. t = 1; Training base classifier: using the first m Each base classifier on the dataset D 1. Perform stacked ensemble training to train a single base classifier. h m1 ( x, l Predicting patient outcomes; Weight update: Calculate the Hamming loss of the base classifier, i.e., the proportion of misclassified labels. e t Calculate and update weights ,use Calculate the next iteration update Repeated integration iteration: Set the number of iterations t = t +1, until the preset number of iterations is reached; Weighted voting in a learning classifier: t = 1,…, T A single classifier h mt Weighted voting is used to improve the learning classifier. h m .

2. The disease prediction method based on causal inference and dynamic integration of multi-labels according to claim 1, characterized in that, The demographic information includes: gender, age, height, and weight; lifestyle information includes: smoking history and drinking history; physical examination information includes: biochemical tests, electrocardiogram information, and imaging data; and past medical history information includes: the patient's personal medical history and family medical history.

3. The disease prediction method based on causal inference and dynamic integration of multi-labels according to claim 1, characterized in that, The steps of dynamically constructing new multi-label ensemble prediction models by combining different numbers and types of base classifiers include: Initialization: The original dataset with causal effects D s It is an empty set; Classify the training samples: Use the trained base learning classifier h m ( x , l ) features x 1. Classify and obtain c 1m = h m ( x 1, l ); Update dataset D s : D s = { c 11 , c 12 ,…, c 1m , y }, and repeat the process of classifying the training samples until the correct classification is achieved. N All hospitalized patients were classified and predicted. c nm = h m (x n , l) To obtain a new dataset D s ={(( c i1 , c i2 ,…, c im ), y )}; The ensemble part of training the model: using a new dataset D s The model is trained, dynamically selected based on the base learner pool, and then stacked and ensembled. This part uses a new learning algorithm. Z ,but H = Z ( D s ); Output: H ( x )= H ( h 1( x, l ) , h 2( x ,l ) ,…,h m ( x, l )).

4. The disease prediction method based on causal inference and dynamic integration of multi-labels according to claim 3, characterized in that, After obtaining multi-source information about the patient, it also includes: The multi-source information is preprocessed and cleaned to remove noise and outliers, and feature selection and dimensionality reduction are performed.

5. A disease prediction system based on causal inference and dynamic integration of multi-labels, characterized in that, include: The data collection module is used to acquire multi-source information about patients, including: demographic information, lifestyle information, physical examination information, chief complaints, past medical history and past medication information; The causal inference module is used to analyze the causal relationships between various features to build a causal model, and to select a feature set with causal effects based on the causal inference model. The dynamic ensemble multi-label algorithm module is used to train multiple multi-label base learning classifiers using the feature set with causal effects, and update the weights through stacked ensemble to obtain the best-performing prediction model. The disease prediction model determination module is used to dynamically construct new multi-label integrated prediction models by combining different numbers and types of optimal prediction models, and select the combination model with the highest prediction performance to predict diseases. The causal inference module is also used to construct a Bayesian network: Let Let L be the joint probability distribution of outcome y, y∈L, n = 1,..., N, where N represents the number of patients, L = {l1,l2,...,lq} is a set of q distinct binary outcome labels, and U is the set of nodes G in a directed acyclic graph. If G...<U, G, P(U)> If the Markov condition is satisfied, then<U, G, P(U)> The triples are called Bayesian networks, where each variable is independent of any subset of non-child terms in G, given the parent terms. Training Markov Chains: Setting BN<U, G, P(U)> For loyalty-based hypothesis Bayesian networks A Markov blanket ∈ F is denoted as MB(Fi), where For a unique term, F represents a distinct feature. express The set of parent nodes, i.e., those directly affected The set of variables; express The set of child nodes, i.e. The set of variables that are directly affected; Indicates and The set of other nodes that share the same parent node, and A set of variables that have an indirect influence relationship; Filtering multi-label association features: targeting feature sets probability of the outcome Maximize, where , The causal feature selection for dataset D is defined as follows: Where T represents the possible disease category to be output; Repeat the above process of training Markov chains and filtering multi-label association features, and finally maximize the probability distribution of the most feature corresponding to all outcome labels y∈L, and select the feature set with causal effect; The dynamically integrated multi-tag module is also used for initialization: for all individual patients. i Initialize the weights for each patient. W 1( i, l ), obtain the initial sample dataset D 1, l Represents the label and sets the number of iterations. t = 1; Training base classifier: using the first m Each base classifier on the dataset D 1. Perform stacked ensemble training to train a single base classifier. h m1 ( x, l Predicting patient outcomes; Weight update: Calculate the Hamming loss of the base classifier, i.e., the proportion of misclassified labels. e t Calculate and update weights ,use Calculate the next iteration update Repeated integration iteration: Set the number of iterations t = t +1, until the preset number of iterations is reached; Weighted voting in a learning classifier: t = 1,…, T A single classifier h mt Weighted voting is used to improve the learning classifier. h m .

6. The disease prediction system based on causal inference and dynamic integration of multi-labels according to claim 5, characterized in that, Also includes: The data processing module is used to preprocess and clean the multi-source information, remove noise and outliers, and perform feature selection and dimensionality reduction operations. The interface display module is used to show the disease prediction results to users; The disease prediction model update module includes an incremental learning unit and a transfer learning unit. In the incremental learning unit, the weights of new and old samples are adjusted using a method based on sample importance. In the transfer learning unit, the existing prediction model is used to predict new diseases.

7. An electronic device, characterized in that, include: A memory and a processor are communicatively connected, the memory storing computer instructions, and the processor executing the computer instructions to perform the disease prediction method based on causal inference and dynamic integrated multi-label as described in any one of claims 1-4.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to perform the disease prediction method based on causal inference and dynamic integration of multiple labels as described in any one of claims 1-4.

Images