A postoperative risk early warning method and system based on machine learning

Abstract

The application relates to the technical field of postoperative early warning, and particularly discloses a postoperative risk early warning method and system based on machine learning, which belongs to the technical field of postoperative early warning, and the method comprises the following steps: acquiring historical patient data of a plurality of patients who have undergone a to-be-early-warned operation, determining an adverse prognosis set of the to-be-early-warned operation and adverse data of each adverse prognosis in the adverse prognosis set; determining an influence factor vector of each adverse prognosis in the adverse prognosis set and an independent prognosis factor set; determining a plurality of preset non-independent factor sets of each adverse prognosis in the adverse prognosis set; determining a non-independent prognosis factor set of each adverse prognosis in the adverse prognosis set; and realizing postoperative risk early warning. The postoperative risk early warning method can comprehensively cover the postoperative risk influence dimension, accurately capture data correlation, realize intelligent transformation from data to risk prediction, provide more accurate and comprehensive postoperative risk early warning for the clinic, and improve the scientificity, accuracy and reliability of postoperative risk early warning.

Description

Technical Field

[0001] This invention relates to the field of postoperative early warning technology, and in particular to a postoperative risk early warning method and system based on machine learning. Background Technology

[0002] Historically, postoperative risk assessment has relied heavily on physician experience and a limited set of clinical indicators, such as patient age and underlying medical conditions, resulting in a lack of objectivity and comprehensiveness. While some statistical risk assessment models have emerged with the development of artificial intelligence and medical informatics, they utilize only single-dimensional data and fail to capture the complex relationships between clinical factors. The rise of machine learning technology has led to its introduction into the medical field, but existing machine learning-based postoperative risk warning methods often lack in-depth data mining and scientific classification, resulting in deficiencies in factor selection and failing to meet the needs for precise clinical risk warning.

[0003] Therefore, this invention proposes a postoperative risk warning method and system based on machine learning. Summary of the Invention

[0004] This invention provides a postoperative risk warning method and system based on machine learning. By acquiring and analyzing historical patient data, it identifies a set of adverse prognoses for surgeries requiring warning, along with adverse data for each adverse prognosis within this set. It then determines the influencing factor vector and independent prognostic factor set for each adverse prognosis, as well as multiple preset non-independent factor sets and a set of non-independent prognostic factors for each adverse prognosis. Based on historical patient data, the independent prognostic factor sets, and the non-independent prognostic factor sets for all adverse prognoses, postoperative risk warning is achieved. This method comprehensively covers the dimensions of postoperative risk impact, accurately captures the correlation between preoperative, intraoperative, and postoperative data, and realizes an intelligent transformation from data to risk prediction. It provides more accurate and comprehensive postoperative risk warnings for clinical practice, improving the scientific rigor, accuracy, and reliability of postoperative risk warnings.

[0005] This invention provides a postoperative risk warning method based on machine learning, comprising:

[0006] S1: Obtain historical patient data from multiple patients who have undergone surgery for which warning is pending, analyze the historical patient data to determine the set of adverse prognoses for surgery for which warning is pending, and the adverse data for each adverse prognosis in the set of adverse prognoses;

[0007] S2: Based on historical patient data, the set of adverse prognoses, and the adverse data of each adverse prognosis in the set of adverse prognoses, determine the influencing factor vector and the set of independent prognostic factors for each adverse prognosis in the set of adverse prognoses;

[0008] S3: Based on the influencing factor vector of each adverse prognosis in the adverse prognosis set and the independent prognosis factor set, determine multiple preset non-independent factor sets for each adverse prognosis in the adverse prognosis set;

[0009] S4: Based on the set of all preset non-independent factors for each adverse prognosis in the adverse prognosis set, determine the set of non-independent prognostic factors for each adverse prognosis in the adverse prognosis set;

[0010] S5: Based on historical patient data, the set of independent prognostic factors for all adverse outcomes in the adverse outcome set, and the set of non-independent prognostic factors, postoperative risk warning is achieved.

[0011] Preferably, a machine learning-based postoperative risk warning method acquires historical patient data from multiple patients who have undergone surgeries requiring warning, including:

[0012] Obtain historical patient sub-data for each patient who has undergone surgery requiring early warning within a specified time period. The historical patient sub-data includes patient information data, preoperative physiological data, intraoperative operation data, and postoperative physiological data.

[0013] Historical patient data is determined based on historical patient sub-data of all patients who underwent surgeries requiring early warning within a specified time period.

[0014] Preferably, a machine learning-based postoperative risk warning method analyzes historical patient data to determine a set of poor prognoses for surgeries requiring warning, and poor data for each poor prognosis in the set, including:

[0015] Determine whether there are adverse events in the postoperative physiological data of each patient's historical patient subdata in the historical patient data. If so, extract the adverse data from the postoperative physiological data of each patient's historical patient subdata in the historical patient data. The adverse data includes at least one or more adverse subdata, which includes the event patient, adverse event, time of adverse occurrence, and severity of adverse event.

[0016] The adverse data of all patients were statistically analyzed to determine the set of adverse prognoses for surgeries requiring early warning, and the adverse data for each adverse prognosis in the set of adverse prognoses were determined. The adverse data included multiple adverse sub-data of the corresponding adverse events.

[0017] Preferably, a machine learning-based postoperative risk warning method, based on historical patient data, a set of adverse prognoses, and adverse data for each adverse prognosis in the set of adverse prognoses, determines the influencing factor vector and the set of independent prognostic factors for each adverse prognosis in the set of adverse prognoses, including:

[0018] For each adverse sub-data in the adverse data of each adverse prognosis in the adverse prognosis set, feature extraction is performed to determine the adverse feature vector of each adverse sub-data in the adverse data of each adverse prognosis in the adverse prognosis set;

[0019] Based on the event patients in each sub-data of adverse data of each adverse prognosis in the adverse prognosis set, feature extraction is performed on the patient information data, preoperative physiological data and intraoperative operation data of the corresponding historical patient sub-data in the historical patient data to determine the influencing factor vector of each adverse prognosis in the adverse prognosis set, and the influencing factor value vector of each sub-data of adverse data of each adverse prognosis in the adverse prognosis set.

[0020] Based on the adverse feature vectors, influencing factor value vectors, and Cox proportional hazards model of all adverse subdata in each adverse prognosis in the adverse prognosis set, the first independent prognostic factor subset for each adverse prognosis in the adverse prognosis set is determined, and the number of first influencing factors in the first independent prognostic factor subset is recorded.

[0021] Pearson correlation analysis was performed on the adverse feature vectors and influencing factor value vectors of all adverse subdata in each adverse prognosis set to determine the single correlation value between each influencing factor and the adverse prognosis.

[0022] For each adverse prognosis in the adverse prognosis set, sort all the single correlation values ​​between the influencing factors and the adverse prognosis in the influencing factor vector of each adverse prognosis from largest to smallest, and select the first number of influencing factors after sorting to determine the second independent prognostic factor subset for each adverse prognosis in the adverse prognosis set;

[0023] Based on the first and second subsets of independent prognostic factors for each adverse prognosis in the adverse prognosis set, the set of independent prognostic factors for each adverse prognosis in the adverse prognosis set is determined, wherein the set of independent prognostic factors includes multiple influencing factors.

[0024] Preferably, a postoperative risk warning method based on machine learning, comprising: a vector of influencing factors for each adverse prognosis in a set of adverse prognoses and a set of independent prognostic factors; and determining multiple preset sets of non-independent factors for each adverse prognosis in the set of adverse prognoses, including:

[0025] Remove all influencing factors from the independent prognostic factor set from the influencing factor vector of each adverse prognosis in the adverse prognosis set, and determine the non-independent factor candidate set for each adverse prognosis in the adverse prognosis set. The non-independent factor candidate set includes multiple influencing factors other than the independent prognostic factor set.

[0026] Cluster analysis was performed on all influencing factors in the candidate set of non-independent factors for each adverse prognosis in the adverse prognosis set. Based on the cluster analysis results, multiple cluster sets were determined, and the cluster similarity and the number of second influencing factors in each cluster set were recorded.

[0027] Based on the cluster similarity of each cluster of each poor prognosis in the poor prognosis set and the number of second influencing factors, the number of preset factors in each cluster is calculated.

[0028] For each cluster of adverse prognosis in the adverse prognosis set, all influencing factors are sorted from largest to smallest based on their single correlation value with the adverse prognosis. The top number of influencing factors after sorting each cluster is selected as the preset non-independent factor set for each cluster.

[0029] Based on the preset set of non-independent factors of all clusters of each adverse prognosis in the adverse prognosis set, multiple preset sets of non-independent factors for each adverse prognosis in the adverse prognosis set are determined.

[0030] Preferably, a postoperative risk warning method based on machine learning determines the set of independent prognostic factors for each adverse prognosis in the adverse prognosis set, based on a pre-defined set of independent factors for each adverse prognosis in the adverse prognosis set, including:

[0031] Based on each preset non-independent factor set for each adverse prognosis in the adverse prognosis set, and the single correlation values ​​between all influencing factors and adverse prognosis in each preset non-independent factor set, the optimal subset of each preset non-independent factor set for each adverse prognosis in the adverse prognosis set is selected;

[0032] For each adverse prognosis set, extract all the optimal subsets of the preset non-independent factor set that are not empty. Based on all the extracted optimal subsets, determine the non-independent prognostic factor set for each adverse prognosis set.

[0033] Preferably, a machine learning-based postoperative risk warning method, based on historical patient data, a set of independent prognostic factors for all adverse prognoses in the adverse prognostic set, and a set of independent prognostic factors, achieves postoperative risk warning, including:

[0034] Based on the set of independent prognostic factors and the set of non-independent prognostic factors for all adverse prognoses in the adverse prognostic set, a postoperative risk assessment and early warning model is constructed.

[0035] The preoperative physiological data and intraoperative operation data of all historical patient subdata in the historical patient data are used as input to the postoperative risk warning model, and the postoperative physiological data of all historical patient subdata in the historical patient data are used as output to train the postoperative risk warning model.

[0036] Based on a well-trained postoperative risk assessment and early warning model, postoperative risk warnings for patients can be achieved.

[0037] This invention provides a machine learning-based postoperative risk warning system for executing any one of the machine learning-based postoperative risk warning methods in Examples 1 to 7, comprising:

[0038] Acquisition module: Acquires historical patient data of multiple patients who have undergone surgery to be alerted, analyzes the historical patient data to determine the set of adverse prognoses for surgery to be alerted, and the adverse data for each adverse prognosis in the set of adverse prognoses;

[0039] Independent prognostic module: Based on historical patient data, the set of poor prognoses, and the adverse data of each poor prognosis in the set of poor prognoses, determine the influencing factor vector of each poor prognosis in the set of poor prognoses and the set of independent prognostic factors;

[0040] Preset non-independent module: Based on the influencing factor vector of each adverse prognosis in the adverse prognosis set and the independent prognosis factor set, determine multiple preset non-independent factor sets for each adverse prognosis in the adverse prognosis set;

[0041] Non-independent prognostic module: Based on the set of all preset non-independent factors for each adverse prognosis in the adverse prognosis set, determine the set of non-independent prognostic factors for each adverse prognosis in the adverse prognosis set;

[0042] Early warning module: Based on historical patient data, the set of independent prognostic factors for all adverse prognoses in the adverse prognostic set, and the set of non-independent prognostic factors, it realizes postoperative risk early warning.

[0043] The beneficial effects of this invention compared to existing technologies are as follows: By acquiring and analyzing historical patient data, it identifies the set of adverse prognoses for surgeries requiring early warning, as well as the adverse data for each adverse prognosis within this set. It also identifies the influencing factor vector and independent prognostic factor set for each adverse prognosis, multiple preset non-independent factor sets for each adverse prognosis, and a set of non-independent prognostic factors for each adverse prognosis. Based on historical patient data, the independent prognostic factor sets for all adverse prognoses, and the non-independent prognostic factor sets, it achieves postoperative risk early warning. This comprehensively covers the dimensions of postoperative risk impact, accurately captures the correlation between preoperative, intraoperative, and postoperative data, and realizes an intelligent transformation from data to risk prediction. It provides more accurate and comprehensive postoperative risk early warning for clinical practice, improves the scientific rigor, accuracy, and reliability of postoperative risk early warning, promotes the construction of high-quality industry datasets in the medical field, and lays the foundation for building large-scale model data products for the medical industry.

[0044] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in this application.

[0045] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0046] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0047] Figure 1 This is a flowchart of a postoperative risk warning method based on machine learning in an embodiment of the present invention;

[0048] Figure 2 This is a schematic diagram of a postoperative risk warning system based on machine learning, as described in an embodiment of the present invention. specific Application method

[0049] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention. Detailed Implementation

[0050] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention. Example 1

[0051] This invention provides a postoperative risk warning method based on machine learning, referring to... Figure 1 ,include:

[0052] S1: Obtain historical patient data from multiple patients who have undergone surgery for which warning is pending, analyze the historical patient data to determine the set of adverse prognoses for surgery for which warning is pending, and the adverse data for each adverse prognosis in the set of adverse prognoses;

[0053] S2: Based on historical patient data, the set of adverse prognoses, and the adverse data of each adverse prognosis in the set of adverse prognoses, determine the influencing factor vector and the set of independent prognostic factors for each adverse prognosis in the set of adverse prognoses;

[0054] S3: Based on the influencing factor vector of each adverse prognosis in the adverse prognosis set and the independent prognosis factor set, determine multiple preset non-independent factor sets for each adverse prognosis in the adverse prognosis set;

[0055] S4: Based on the set of all preset non-independent factors for each adverse prognosis in the adverse prognosis set, determine the set of non-independent prognostic factors for each adverse prognosis in the adverse prognosis set;

[0056] S5: Based on historical patient data, the set of independent prognostic factors for all adverse outcomes in the adverse outcome set, and the set of non-independent prognostic factors, postoperative risk warning is achieved.

[0057] In this embodiment, historical data from multiple patients who have undergone surgeries requiring early warning are first collected from the medical system. This data includes patient information, pre- and post-operative physiological data, and intraoperative operational data. By analyzing the post-operative physiological data, it is determined whether adverse events (such as infection or bleeding) occur. If so, adverse data containing information such as the patient involved, the adverse event, the time of occurrence, and its severity are extracted. These adverse prognostic sets and corresponding adverse data are then statistically summarized, laying the foundation for subsequent analysis.

[0058] In this embodiment, features are extracted from adverse subdata within the adverse data to form an adverse feature vector. Based on the patients involved in the event, patient information, preoperative and intraoperative data features are extracted from historical patient data to determine the influencing factor vector and influencing factor value vector. Using the Cox proportional hazards model and Pearson correlation analysis, a set of independent prognostic factors that have a significant impact on adverse prognosis is selected.

[0059] In this embodiment, independent prognostic factors are removed from the influencing factor vector to obtain a candidate set of non-independent factors. Cluster analysis is then performed on these factors, and a preset number of factors is calculated based on cluster similarity and the number of factors. These factors are then sorted according to their single correlation value with poor prognosis, and the top preset number of factors are selected to form a preset set of non-independent factors for each cluster.

[0060] In this embodiment, based on the single correlation value between each factor and adverse prognosis in the preset set of non-independent factors, the optimal subset of each set is selected, the non-empty optimal subset is extracted and integrated to form a set of non-independent prognostic factors.

[0061] In this embodiment, historical patient data, a set of independent prognostic factors, and a set of non-independent prognostic factors are used to construct a postoperative risk warning model. The model is trained using preoperative and intraoperative data as input and postoperative data as output, and finally, postoperative risk warning for patients is achieved based on the trained model.

[0062] The beneficial effects of the above technologies are as follows: By acquiring and analyzing historical patient data, the set of adverse prognoses for surgeries requiring early warning is determined, along with the adverse data for each adverse prognosis within this set. The influencing factor vector and independent prognostic factor set for each adverse prognosis within the set are determined, as well as multiple pre-defined sets of non-independent factors for each adverse prognosis within the set, and the set of non-independent prognostic factors for each adverse prognosis within the set. Based on historical patient data, the sets of independent and non-independent prognostic factors for all adverse prognoses in the set, postoperative risk warnings are achieved. This comprehensively covers the dimensions of postoperative risk impact, accurately captures the correlation between preoperative, intraoperative, and postoperative data, and realizes an intelligent transformation from data to risk prediction. It provides more accurate and comprehensive postoperative risk warnings for clinical practice, improving the scientific rigor, accuracy, and reliability of postoperative risk warnings. Example 2

[0063] Based on Example 1, a machine learning-based postoperative risk warning method acquires historical patient data from multiple patients who have undergone surgeries requiring warning, including:

[0064] Obtain historical patient sub-data for each patient who has undergone surgery requiring early warning within a specified time period. The historical patient sub-data includes patient information data, preoperative physiological data, intraoperative operation data, and postoperative physiological data.

[0065] Historical patient data is determined based on historical patient sub-data of all patients who underwent surgeries requiring early warning within a specified time period.

[0066] In this embodiment, the time range for data collection is clearly defined to avoid data interference caused by changes in medical technology and treatment standards over a long period, ensuring data homogeneity and timeliness. For example, data from the past three years is selected because the surgical techniques and anesthesia methods used during that time are more consistent with the current medical environment and can effectively reflect the true situation of current surgeries.

[0067] In this embodiment, multi-dimensional data is collected for each patient who has undergone a pre-examination surgery. Patient information data includes age, gender, underlying diseases, family medical history, etc., which reflect the individual characteristics of the patient; preoperative physiological data includes indicators such as blood routine, electrocardiogram, liver and kidney function, reflecting the patient's physical function before surgery; intraoperative operation data records the operation time, blood loss, anesthesia method, surgical instruments used, etc., presenting details of the surgical process; postoperative physiological data includes vital signs, wound healing status, occurrence of complications, etc., showing the postoperative recovery status.

[0068] In this embodiment, historical patient sub-data for all patients within a specified time period is aggregated and integrated to form a complete historical patient dataset. This process systematizes scattered individual data, constructs a structured database, and provides a comprehensive and unified data foundation for subsequent analysis and model training.

[0069] The beneficial effects of the above technologies are: obtaining historical patient data from multiple patients who have undergone pre-examination surgeries can improve data integrity and provide data support for determining the set of poor prognoses and the adverse data of each poor prognosis in the set of poor prognoses. Example 3

[0070] Building upon Example 2, a machine learning-based postoperative risk warning method analyzes historical patient data to determine a set of poor prognoses for surgeries requiring warning, and adverse data for each poor prognosis within this set, including:

[0071] Determine whether there are adverse events in the postoperative physiological data of each patient's historical patient subdata in the historical patient data. If so, extract the adverse data from the postoperative physiological data of each patient's historical patient subdata in the historical patient data. The adverse data includes at least one or more adverse subdata, which includes the event patient, adverse event, time of adverse occurrence, and severity of adverse event.

[0072] The adverse data of all patients were statistically analyzed to determine the set of adverse prognoses for surgeries requiring early warning, and the adverse data for each adverse prognosis in the set of adverse prognoses were determined. The adverse data included multiple adverse sub-data of the corresponding adverse events.

[0073] In this embodiment, the postoperative physiological data of each patient's historical patient data is first reviewed one by one. Postoperative physiological data contains a wealth of crucial information, such as vital signs (body temperature, pulse, respiration, blood pressure, etc.), laboratory test indicators (complete blood count, blood biochemistry, coagulation function, etc.), and the functional status of various bodily systems. This data reflects the patient's actual physical condition after surgery.

[0074] In this embodiment, during the review process, it is determined whether there are any adverse events. Adverse events refer to those situations that negatively affect the patient's postoperative recovery, such as postoperative infection (like wound infection, lung infection, etc.), bleeding (surgical site bleeding, internal bleeding, etc.), organ dysfunction (such as heart failure, kidney failure, etc.), thrombosis (deep vein thrombosis, pulmonary embolism, etc.), and neurological complications (such as cognitive impairment, nerve damage, etc.).

[0075] In this embodiment, once an adverse event is detected, relevant adverse data is extracted from the patient's postoperative physiological data. The adverse data consists of multiple sub-data items. The "Event Patient" section identifies the specific patient involved in the adverse event, facilitating subsequent tracking and analysis of individual cases. The "Adverse Event" section details the specific adverse condition that occurred, the "Adverse Onset Time" records the time when the adverse event began, and the "Adverse Severity" section quantifies the severity of the adverse event, such as the severity of infection, the specific degree of organ dysfunction, and the number of complications, in order to measure its impact on the patient's health.

[0076] In this embodiment, adverse data extracted from all patients is collected and summarized. This adverse data may come from different patients and covers various types of adverse events and their corresponding detailed information.

[0077] In this embodiment, statistical analysis is performed on the aggregated adverse data. By analyzing the type, frequency, and severity of adverse events, similar adverse events are categorized and organized to determine the set of adverse prognoses for surgeries requiring early warning. The set of adverse prognoses is a collection of various adverse outcomes that may occur in this type of surgery.

[0078] In this embodiment, for each adverse prognosis in the adverse prognosis set, its corresponding adverse data is further clarified. For each specific adverse prognosis, multiple related adverse sub-data are recorded in detail. This data can describe the characteristics and situation of the adverse prognosis in more detail, including the patients involved, the specific manifestations of the adverse event, the time of occurrence, and the severity.

[0079] The beneficial effects of the above technologies are: analyzing historical patient data to determine the set of adverse prognoses for surgeries requiring early warning, as well as the adverse data for each adverse prognosis in the set, can lay a solid data foundation for subsequent in-depth analysis of the influencing factors of adverse prognoses and the realization of accurate postoperative risk warning. Example 4

[0080] Building upon Example 3, a machine learning-based postoperative risk warning method, based on historical patient data, a set of adverse prognoses, and adverse data for each adverse prognosis within the set, determines the influencing factor vector and the set of independent prognostic factors for each adverse prognosis within the set, including:

[0081] For each adverse sub-data in the adverse data of each adverse prognosis in the adverse prognosis set, feature extraction is performed to determine the adverse feature vector of each adverse sub-data in the adverse data of each adverse prognosis in the adverse prognosis set;

[0082] Based on the event patients in each sub-data of adverse data of each adverse prognosis in the adverse prognosis set, feature extraction is performed on the patient information data, preoperative physiological data and intraoperative operation data of the corresponding historical patient sub-data in the historical patient data to determine the influencing factor vector of each adverse prognosis in the adverse prognosis set, and the influencing factor value vector of each sub-data of adverse data of each adverse prognosis in the adverse prognosis set.

[0083] Based on the adverse feature vectors, influencing factor value vectors, and Cox proportional hazards model of all adverse subdata in each adverse prognosis in the adverse prognosis set, the first independent prognostic factor subset for each adverse prognosis in the adverse prognosis set is determined, and the number of first influencing factors in the first independent prognostic factor subset is recorded.

[0084] Pearson correlation analysis was performed on the adverse feature vectors and influencing factor value vectors of all adverse subdata in each adverse prognosis set to determine the single correlation value between each influencing factor and the adverse prognosis.

[0085] For each adverse prognosis in the adverse prognosis set, sort all the single correlation values ​​between the influencing factors and the adverse prognosis in the influencing factor vector of each adverse prognosis from largest to smallest, and select the first number of influencing factors after sorting to determine the second independent prognostic factor subset for each adverse prognosis in the adverse prognosis set;

[0086] Based on the first and second subsets of independent prognostic factors for each adverse prognosis in the adverse prognosis set, the set of independent prognostic factors for each adverse prognosis in the adverse prognosis set is determined, wherein the set of independent prognostic factors includes multiple influencing factors.

[0087] In this embodiment, feature extraction is performed on each sub-data item of adverse prognosis within each adverse prognosis set. For example, features such as the type of event (e.g., infection, bleeding), nature (acute or chronic), specific numerical value of adverse time, and time interval are extracted from the adverse sub-data. In this way, an adverse feature vector is determined for each adverse sub-data item. This vector contains the key feature information of that adverse sub-data item for subsequent analysis.

[0088] In this embodiment, for each adverse prognosis sub-data within the adverse prognosis data of each adverse prognosis in the adverse prognosis set, corresponding historical patient sub-data is found in the historical patient data. Feature extraction is performed on patient information data (such as age, gender, underlying diseases, etc.), preoperative physiological data (such as various physiological indicators, examination results, etc.), and intraoperative operation data (such as surgical method, operation duration, blood loss, etc.) from these corresponding historical patient sub-data. After feature extraction, an influencing factor vector is determined for each adverse prognosis in the adverse prognosis set. This vector contains various influencing factors that may be related to the adverse prognosis. Simultaneously, for each adverse sub-data, a corresponding influencing factor value vector is determined. This vector represents the specific value or degree of influence of each influencing factor in the context of that adverse sub-data.

[0089] In this embodiment, the adverse feature vectors and influencing factor value vectors of all adverse sub-data in each adverse prognosis in the adverse prognosis set are comprehensively considered. The Cox proportional hazards model is used to analyze and calculate which factors have a significant independent impact on the adverse prognosis. This results in the first independent prognostic factor subset for each adverse prognosis in the adverse prognosis set, and the number of influencing factors in this subset is recorded, i.e., the number of first influencing factors.

[0090] In this embodiment, the Cox proportional hazards model is a statistical model used to analyze survival data, and here it is used to assess the degree of risk impact of various factors on adverse prognosis.

[0091] In this embodiment, Pearson correlation analysis is performed on the adverse feature vectors and influencing factor value vectors of all adverse subdata in each adverse prognosis set.

[0092] In this embodiment, the Pearson correlation coefficient is an indicator that measures the degree of linear correlation between two variables, with a value ranging from -1 to 1. Through this analysis, the single correlation value between each influencing factor in the influencing factor vector and the adverse prognosis is determined, thereby measuring the degree of linear association between each influencing factor and the adverse prognosis.

[0093] In this embodiment, all the single correlation values ​​between the influencing factors and the adverse prognosis in the influencing factor vector of each adverse prognosis in the adverse prognosis set are sorted in descending order. Based on the number of the first influencing factors recorded previously, the corresponding number of influencing factors at the top of the sorting are selected. These factors form the second independent prognostic factor subset for each adverse prognosis in the adverse prognosis set.

[0094] In this embodiment, the first independent prognostic factor subset and the second independent prognostic factor subset of each adverse prognosis in the adverse prognosis set are integrated. The integrated set is the independent prognostic factor set of each adverse prognosis in the adverse prognosis set. This set contains multiple factors that have a significant independent impact on the adverse prognosis.

[0095] The beneficial effects of the above technologies are as follows: Based on historical patient data, the set of adverse prognoses, and the adverse data of each adverse prognosis in the set of adverse prognoses, the influencing factor vector and the set of independent prognostic factors for each adverse prognosis in the set of adverse prognoses can be determined. This can comprehensively explore the key independent factors affecting adverse prognoses, avoid the omission of factors, provide a more accurate and reliable basis for postoperative risk warning, and improve the accuracy and effectiveness of risk assessment. Example 5

[0096] Building upon Example 4, a machine learning-based postoperative risk warning method determines multiple preset non-independent factor sets for each adverse prognosis in the adverse prognosis set, based on the influencing factor vector of each adverse prognosis in the adverse prognosis set and the set of independent prognostic factors, including:

[0097] Remove all influencing factors from the independent prognostic factor set from the influencing factor vector of each adverse prognosis in the adverse prognosis set, and determine the non-independent factor candidate set for each adverse prognosis in the adverse prognosis set. The non-independent factor candidate set includes multiple influencing factors other than the independent prognostic factor set.

[0098] Cluster analysis was performed on all influencing factors in the candidate set of non-independent factors for each adverse prognosis in the adverse prognosis set. Based on the cluster analysis results, multiple cluster sets were determined, and the cluster similarity and the number of second influencing factors in each cluster set were recorded.

[0099] Based on the cluster similarity of each cluster of each poor prognosis in the poor prognosis set and the number of second influencing factors, the number of preset factors in each cluster is calculated.

[0100] For each cluster of adverse prognosis in the adverse prognosis set, all influencing factors are sorted from largest to smallest based on their single correlation value with the adverse prognosis. The top number of influencing factors after sorting each cluster is selected as the preset non-independent factor set for each cluster.

[0101] Based on the preset set of non-independent factors of all clusters of each adverse prognosis in the adverse prognosis set, multiple preset sets of non-independent factors for each adverse prognosis in the adverse prognosis set are determined.

[0102] In this embodiment, the influencing factor vector for each adverse prognosis in the known adverse prognosis set encompasses all factors that may affect that adverse prognosis, while the independent prognostic factor set contains key factors that have a direct and independent impact on the adverse prognosis. Based on this, all factors in the independent prognostic factor set are removed from the influencing factor vector. For example, if age or underlying disease is an independent prognostic factor, it is removed from the influencing factor vector, and the remaining factors form a candidate set of non-independent factors, focusing on non-independent influencing factors that may have synergistic effects.

[0103] In this embodiment, cluster analysis is performed on all influencing factors in the candidate set of non-independent factors, using algorithms such as K-means and hierarchical clustering. Factors are divided into different cluster sets based on the similarity of data features and clinical relevance among them. For example, factors related to surgical trauma recovery (such as surgical incision size and intraoperative tissue damage) are clustered together. These cluster sets may be based on physiological function, including influencing factors such as heart rate and abnormal electrocardiogram indicators, or they may be based on disease risk, including influencing factors such as hyperlipidemia and hyperglycemia. Simultaneously, the cluster similarity of each cluster set is recorded to measure the degree of correlation between factors within each cluster; the number of secondary influencing factors, i.e., the total number of factors within each cluster set, is also recorded to provide a basis for subsequent screening.

[0104] In this embodiment, the formula for calculating the preset number of factors for each cluster set based on the cluster similarity of each cluster set and the number of second influencing factors can be expressed as:

[0105] ;

[0106] in, This represents the number of preset factors in the b-th cluster set for the a-th adverse prognosis. This represents the similarity between the i-th and j-th influencing factors in the b-th cluster set for the a-th adverse prognosis. This represents the number of second influencing factors in the b-th cluster set of the a-th adverse prognosis. This represents the cluster similarity of the b-th cluster set for the a-th adverse prognosis. Let w1 represent the similarity standard deviation of the b-th cluster set for the a-th poor prognosis, w1 represent the similarity coefficient, β represent the discrete penalty coefficient, and γ represent the size adjustment coefficient. This indicates rounding up to the nearest integer.

[0107] In this embodiment, the similarity coefficient w1 ranges from 0 to 1, and can be 0.2; the discrete penalty coefficient β ranges from 0 to 1, and can be 1.5; and the scale adjustment coefficient γ ranges from 0 to 1, and can be 0.8.

[0108] In this embodiment, all influencing factors in each cluster are sorted from largest to smallest based on their single correlation value with adverse prognosis (obtained through Pearson correlation analysis). The factors ranked highest are selected according to the calculated preset number of factors to form a preset set of non-independent factors for each cluster.

[0109] In this embodiment, the preset non-independent factor sets of all clusters under the same adverse prognosis are aggregated to form multiple preset non-independent factor sets for each adverse prognosis in the adverse prognosis set. Each preset non-independent factor set represents a class of non-independent factor combinations with similarity.

[0110] In this embodiment, the order of all influencing factors in the preset set of non-independent factors is still sorted according to the size of their corresponding single correlation.

[0111] In this embodiment, each cluster of each poor prognosis in the poor prognosis set corresponds to a preset non-independent factor set.

[0112] The beneficial effects of the above technology are as follows: Based on the influencing factor vector of each adverse prognosis in the adverse prognosis set and the set of independent prognostic factors, multiple preset non-independent factor sets of each adverse prognosis in the adverse prognosis set can be determined. This can more comprehensively capture the synergistic effect of non-independent factors, provide more detailed and diverse data support for postoperative risk warning, and significantly improve the comprehensiveness and accuracy of risk assessment. Example 6

[0113] Based on Example 4, a postoperative risk warning method based on machine learning determines the set of non-independent prognostic factors for each adverse prognosis in the adverse prognosis set, based on all preset non-independent factor sets for each adverse prognosis in the adverse prognosis set, including:

[0114] Based on each preset non-independent factor set for each adverse prognosis in the adverse prognosis set, and the single correlation values ​​between all influencing factors and adverse prognosis in each preset non-independent factor set, the optimal subset of each preset non-independent factor set for each adverse prognosis in the adverse prognosis set is selected;

[0115] For each adverse prognosis set, extract all the optimal subsets of the preset non-independent factor set that are not empty. Based on all the extracted optimal subsets, determine the non-independent prognostic factor set for each adverse prognosis set.

[0116] In this embodiment, the optimal subset of each preset non-independent factor set for each adverse prognosis in the adverse prognosis set, and the single correlation values ​​between all influencing factors and adverse prognosis in each preset non-independent factor set, is selected using the following formula:

[0117] ;

[0118] in, Let represent the optimal subset of the predefined set of non-independent factors for the b-th cluster set of the a-th adverse prognosis. The subset of the b-th cluster set representing the a-th adverse prognosis Multiple correlation values, ST represents the subset of the first t influencing factors of the predefined non-independent factor set for the b-th cluster of the a-th adverse prognosis, ST represents the predefined multi-correlation threshold, and t represents the subset. The number of influencing factors This represents the penalty coefficient based on the number of influencing factors in the subset. Representing a subset The single correlation value between the c-th influencing factor and the a-th adverse prognosis. This represents the clustering quality adjustment coefficient based on the average similarity of the b-th cluster set with the a-th poor prognosis. This represents the empty set.

[0119] In this embodiment, for each preset set of non-independent factors in the set of adverse prognoses, the factors are filtered using the single correlation values ​​(an indicator measuring the linear correlation between a factor and an adverse prognosis) of all influencing factors and the adverse prognosis. Specifically, the factors are sorted from largest to smallest by the absolute value of their single correlation values. A threshold is set (e.g., only factors with absolute values ​​higher than a certain threshold are retained) or a certain number of top-ranking factors are selected. Factors with weaker influence on the adverse prognosis are removed, thus obtaining the optimal subset of each preset set of non-independent factors, ensuring that the factors in the subset have a strong influence and correlation with the adverse prognosis.

[0120] In this embodiment, the optimal subset of all preset non-independent factor sets for each adverse prognosis in the adverse prognosis set is checked, and cases where the set is empty are eliminated (i.e., all factors in the preset set do not meet the correlation with the adverse prognosis).

[0121] In this embodiment, the remaining non-empty optimal subsets are integrated to form a set of non-independent prognostic factors for each adverse prognosis in the adverse prognosis set. This set integrates key non-independent factors selected from multiple preset sets.

[0122] The beneficial effects of the above technology are as follows: Based on the set of all preset non-independent factors for each adverse prognosis in the adverse prognosis set, the set of non-independent prognostic factors for each adverse prognosis in the adverse prognosis set can be determined, which can improve the scientificity and effectiveness of non-independent factor screening, more accurately reveal the synergistic effect between factors, provide a more reliable basis for postoperative risk warning, and enhance the accuracy of risk assessment and clinical application value. Example 7

[0123] Based on Example 1, a machine learning-based postoperative risk warning method is proposed. This method utilizes historical patient data, a set of independent prognostic factors for all adverse outcomes within a set of adverse outcomes, and a set of independent prognostic factors to achieve postoperative risk warning. The method includes:

[0124] Based on the set of independent prognostic factors and the set of non-independent prognostic factors for all adverse prognoses in the adverse prognostic set, a postoperative risk assessment and early warning model is constructed.

[0125] The preoperative physiological data and intraoperative operation data of all historical patient subdata in the historical patient data are used as input to the postoperative risk warning model, and the postoperative physiological data of all historical patient subdata in the historical patient data are used as output to train the postoperative risk warning model.

[0126] Based on a well-trained postoperative risk assessment and early warning model, postoperative risk warnings for patients can be achieved.

[0127] In this embodiment, the core elements are the set of independent prognostic factors and the set of non-independent prognostic factors for all poor prognoses within the poor prognostic set. Machine learning algorithms (such as logistic regression, random forest, etc.) or deep learning models (such as neural networks) are used to build a model framework, design the model structure and parameters, enabling the model to process data from different dimensions and analyze the comprehensive impact of factors on postoperative risk.

[0128] In this embodiment, preoperative physiological data and intraoperative operands from all historical patient subdata in the historical patient data are used as model input data. Postoperative physiological data are used as output labels. The model is iteratively trained using a large amount of historical data, adjusting model parameters and optimizing model performance to enable it to accurately learn the mapping relationship between preoperative, intraoperative data and postoperative outcomes.

[0129] In this embodiment, the preoperative physiological data and intraoperative operation data of the new patient are input into the trained model. Based on the learned patterns, the model analyzes and processes the input data to predict the adverse prognosis that the patient may have after surgery.

[0130] The beneficial effects of the above technologies are as follows: based on historical patient data, the set of independent prognostic factors for all adverse prognoses in the adverse prognostic set, and the set of non-independent prognostic factors, postoperative risk warning can be achieved. It can comprehensively cover the dimensions of postoperative risk impact, accurately capture the correlation between preoperative, intraoperative, and postoperative data, realize the intelligent transformation from data to risk prediction, provide more accurate and comprehensive postoperative risk warning for clinical practice, and significantly improve the scientificity and effectiveness of risk prevention and control. Example 8

[0131] This invention provides a machine learning-based postoperative risk warning system for executing any of the machine learning-based postoperative risk warning methods in Examples 1 to 7, with reference to... Figure 2 ,include:

[0132] Acquisition module: Acquires historical patient data of multiple patients who have undergone surgery to be alerted, analyzes the historical patient data to determine the set of adverse prognoses for surgery to be alerted, and the adverse data for each adverse prognosis in the set of adverse prognoses;

[0133] Independent prognostic module: Based on historical patient data, the set of poor prognoses, and the adverse data of each poor prognosis in the set of poor prognoses, determine the influencing factor vector of each poor prognosis in the set of poor prognoses and the set of independent prognostic factors;

[0134] Preset non-independent module: Based on the influencing factor vector of each adverse prognosis in the adverse prognosis set and the independent prognosis factor set, determine multiple preset non-independent factor sets for each adverse prognosis in the adverse prognosis set;

[0135] Non-independent prognostic module: Based on the set of all preset non-independent factors for each adverse prognosis in the adverse prognosis set, determine the set of non-independent prognostic factors for each adverse prognosis in the adverse prognosis set;

[0136] Early warning module: Based on historical patient data, the set of independent prognostic factors for all adverse prognoses in the adverse prognostic set, and the set of non-independent prognostic factors, it realizes postoperative risk early warning.

[0137] The beneficial effects of the above technologies are as follows: By acquiring and analyzing historical patient data, the set of adverse prognoses for surgeries requiring early warning is determined, along with the adverse data for each adverse prognosis within this set. The influencing factor vector and independent prognostic factor set for each adverse prognosis within the set are determined, as well as multiple pre-defined sets of non-independent factors for each adverse prognosis within the set, and the set of non-independent prognostic factors for each adverse prognosis within the set. Based on historical patient data, the sets of independent and non-independent prognostic factors for all adverse prognoses in the set, postoperative risk warnings are achieved. This comprehensively covers the dimensions of postoperative risk impact, accurately captures the correlation between preoperative, intraoperative, and postoperative data, and realizes an intelligent transformation from data to risk prediction. It provides more accurate and comprehensive postoperative risk warnings for clinical practice, improving the scientific rigor, accuracy, and reliability of postoperative risk warnings.

[0138] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for postoperative risk early warning based on machine learning, characterized in that, The method comprises the following steps: S1: obtaining historical patient data of a plurality of patients who have undergone the surgery to be warned, analyzing the historical patient data to determine a set of adverse prognoses of the surgery to be warned, and adverse data of each adverse prognosis in the set of adverse prognoses; S2: determining an influence factor vector of each adverse prognosis in the set of adverse prognoses and a set of independent prognostic factors based on the historical patient data, the set of adverse prognoses, and the adverse data of each adverse prognosis in the set of adverse prognoses; S3: determining a plurality of preset non-independent factor sets of each adverse prognosis in the set of adverse prognoses based on the influence factor vector of each adverse prognosis in the set of adverse prognoses and the set of independent prognostic factors; S4: determining a set of non-independent prognostic factors of each adverse prognosis in the set of adverse prognoses based on all the preset non-independent factor sets of each adverse prognosis in the set of adverse prognoses; S5: realizing postoperative risk warning based on the historical patient data, the set of independent prognostic factors of all the adverse prognoses in the set of adverse prognoses, and the set of non-independent prognostic factors. Based on the historical patient data, the set of adverse prognoses, and the adverse data of each adverse prognosis in the set of adverse prognoses, the method comprises the following steps: performing feature extraction on each adverse sub-data in the adverse data of each adverse prognosis in the set of adverse prognoses to determine an adverse feature vector of each adverse sub-data in the adverse data of each adverse prognosis in the set of adverse prognoses; performing feature extraction on patient information data, preoperative physiological data, and intraoperative operation data of corresponding historical patient sub-data in the historical patient data based on event patients in each adverse sub-data in the adverse data of each adverse prognosis in the set of adverse prognoses to determine an influence factor vector of each adverse prognosis in the set of adverse prognoses and an influence factor value vector of each adverse sub-data in the adverse data of each adverse prognosis in the set of adverse prognoses; determining a first independent prognostic factor sub-set of each adverse prognosis in the set of adverse prognoses based on the adverse feature vectors, the influence factor value vectors, and a Cox proportional hazards model of all the adverse sub-data in the adverse data of each adverse prognosis in the set of adverse prognoses, and recording a first influence factor number of the first independent prognostic factor sub-set; performing Pearson correlation value analysis on the adverse feature vectors and the influence factor value vectors of all the adverse sub-data in the adverse data of each adverse prognosis in the set of adverse prognoses to determine a single correlation value of each influence factor in the influence factor vector and the adverse prognosis; sorting the single correlation values of all the influence factors in the influence factor vector of each adverse prognosis in the set of adverse prognoses from large to small, and selecting a first influence factor number of influence factors after the sorting to determine a second independent prognostic factor sub-set of each adverse prognosis in the set of adverse prognoses; determining a set of independent prognostic factors of each adverse prognosis in the set of adverse prognoses based on the first independent prognostic factor sub-set and the second independent prognostic factor sub-set of each adverse prognosis in the set of adverse prognoses, wherein the set of independent prognostic factors comprises a plurality of influence factors.

2. The machine learning-based postoperative risk early warning method of claim 1, wherein, obtaining historical patient data of a plurality of patients who have undergone the surgery to be warned, including: obtaining historical patient sub-data of each patient who has undergone the surgery to be warned within a specified time period, wherein the historical patient sub-data includes patient information data, preoperative physiological data, intraoperative operation data, and postoperative physiological data; determining the historical patient data based on the historical patient sub-data of all patients who have undergone the surgery to be warned within the specified time period.

3. The machine learning-based postoperative risk early warning method of claim 2, wherein, analyzing the historical patient data to determine a set of adverse prognoses of the surgery to be warned, and adverse data of each adverse prognosis in the set of adverse prognoses, including: determining whether the postoperative physiological data of the historical patient sub-data of each patient in the historical patient data has an adverse event, and if so, extracting adverse data in the postoperative physiological data of the historical patient sub-data of each patient in the historical patient data, wherein the adverse data includes at least one and more adverse sub-data, and the adverse sub-data includes an event patient, an adverse event, an adverse occurrence time, and an adverse degree; statistically analyzing all the extracted adverse data of the patients to determine a set of adverse prognoses of the surgery to be warned, and to determine adverse data of each adverse prognosis in the set of adverse prognoses, wherein the adverse data includes a plurality of adverse sub-data of the corresponding adverse event. 4.The postoperative risk early warning method based on machine learning of claim 1, wherein, determining a plurality of preset non-independent factor sets of each adverse prognosis in the set of adverse prognoses based on an influence factor vector of each adverse prognosis in the set of adverse prognoses and a set of independent prognostic factors, including: eliminating all influence factors in the set of independent prognostic factors from the influence factor vector of each adverse prognosis in the set of adverse prognoses to determine a non-independent factor candidate set of each adverse prognosis in the set of adverse prognoses, wherein the non-independent factor candidate set includes a plurality of influence factors other than the set of independent prognostic factors; performing cluster analysis on all influence factors in the non-independent factor candidate set of each adverse prognosis in the set of adverse prognoses, determining a plurality of cluster sets based on the cluster analysis results, and recording cluster similarity and a second number of influence factors of each cluster set; calculating a preset number of factors of each cluster set based on the cluster similarity and the second number of influence factors of each cluster set of each adverse prognosis in the set of adverse prognoses; sorting all influence factors in each cluster set of each adverse prognosis in the set of adverse prognoses from large to small based on a single correlation value with the adverse prognosis, and selecting a preset number of influence factors in the sorted front of each cluster set as a preset non-independent factor set of each cluster set; determining a plurality of preset non-independent factor sets of each adverse prognosis in the set of adverse prognoses based on the preset non-independent factor sets of all cluster sets of each adverse prognosis in the set of adverse prognoses.

5. The machine learning based post-operative risk alerting method of claim 1, wherein, determining a set of non-independent prognostic factors of each adverse prognosis in the set of adverse prognoses based on all preset non-independent factor sets of each adverse prognosis in the set of adverse prognoses, including: selecting an optimal subset of each preset non-independent factor set of each adverse outcome in the adverse outcome set based on each preset non-independent factor set of each adverse outcome in the adverse outcome set and single correlation values of all influencing factors in each preset non-independent factor set and the adverse outcome; extracting all optimal subsets of all preset non-independent factor sets of each adverse outcome in the adverse outcome set based on the fact that the optimal subset of each preset non-independent factor set of each adverse outcome in the adverse outcome set is not an empty set, and determining the non-independent prognostic factor set of each adverse outcome in the adverse outcome set based on the extracted all optimal subsets.

6. The machine learning based post-operative risk alerting method of claim 1, wherein, Based on the historical patient data, the independent prognostic factor set and the non-independent prognostic factor set of all adverse outcomes in the adverse outcome set, the postoperative risk early warning is realized, including: Based on the independent prognostic factor set and the non-independent prognostic factor set of all adverse outcomes in the adverse outcome set, a postoperative risk early warning model is constructed; The preoperative physiological data and intraoperative operation data of all historical patient sub-data in the historical patient data are taken as the input of the postoperative risk early warning model, and the postoperative physiological data of all historical patient sub-data in the historical patient data are taken as the output of the postoperative risk early warning model. The postoperative risk early warning model is trained. Based on the trained postoperative risk early warning model, the postoperative risk early warning of the patient is realized.

7. A machine learning based post-operative risk alert system, characterized in that, A method for performing any one of claims 1 to 6 based on machine learning postoperative risk early warning, comprising: An acquisition module: acquiring historical patient data of a plurality of patients who have undergone a to-be-early-warned surgery, analyzing the historical patient data to determine an adverse outcome set of the to-be-early-warned surgery, and adverse data of each adverse outcome in the adverse outcome set; An independent prognostic module: based on the historical patient data, the adverse outcome set and the adverse data of each adverse outcome in the adverse outcome set, determining the influencing factor vector of each adverse outcome in the adverse outcome set and the independent prognostic factor set; A preset non-independent module: based on the influencing factor vector of each adverse outcome in the adverse outcome set and the independent prognostic factor set, determining a plurality of preset non-independent factor sets of each adverse outcome in the adverse outcome set; A non-independent prognostic module: based on all preset non-independent factor sets of each adverse outcome in the adverse outcome set, determining the non-independent prognostic factor set of each adverse outcome in the adverse outcome set; An early warning module: based on the historical patient data, the independent prognostic factor set and the non-independent prognostic factor set of all adverse outcomes in the adverse outcome set, realizing postoperative risk early warning; The independent prognostic module comprises: characteristic extraction is performed on each adverse sub-data in the adverse data of each adverse outcome in the adverse outcome set to determine the adverse feature vector of each adverse sub-data in the adverse data of each adverse outcome in the adverse outcome set; Based on the event patients in each adverse sub-data in the adverse data of each adverse prognosis in the adverse prognosis set, feature extraction is performed on the patient information data, preoperative physiological data and intraoperative operation data of the corresponding historical patient sub-data in the historical patient data, the influence factor vector of each adverse prognosis in the adverse prognosis set is determined, and the influence factor value vector of each adverse sub-data in the adverse data of each adverse prognosis in the adverse prognosis set is determined; Based on the adverse feature vectors, influence factor value vectors and Cox proportional hazards model of all adverse sub-data in the adverse data of each adverse prognosis in the adverse prognosis set, a first independent prognosis factor sub-set of each adverse prognosis in the adverse prognosis set is determined, and the first influence factor number of the first independent prognosis factor sub-set is recorded; Pearson correlation value analysis is performed on the adverse feature vectors and influence factor value vectors of all adverse sub-data in the adverse data of each adverse prognosis in the adverse prognosis set, and a single correlation value of each influence factor in the influence factor vector and the adverse prognosis is determined; The single correlation values of all influence factors in the influence factor vector and the adverse prognosis in the adverse prognosis set are sorted from large to small, and the first influence factor number of the sorted influence factors is selected to determine a second independent prognosis factor sub-set of each adverse prognosis in the adverse prognosis set; Based on the first independent prognosis factor sub-set and the second independent prognosis factor sub-set of each adverse prognosis in the adverse prognosis set, an independent prognosis factor set of each adverse prognosis in the adverse prognosis set is determined, wherein the independent prognosis factor set includes multiple influence factors.

Images