Data processing method, training method of sepsis prediction model and intelligent equipment
By using the reverse sequence of time series data and the pre-trained time series prediction model in the sepsis prediction model, the forward and reverse correlation of the data is determined, and the problem of training data selection in the prior art affecting the prediction accuracy is solved, and a more efficient training and prediction effect of the sepsis prediction model is achieved.
Patent Information
- Application Number
- CN202411842582.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-13
- Publication Date
- 2025-04-29
AI Technical Summary
In the prior art, when selecting training data, data with weak correlation with target events is easily introduced, which affects the training effect and accuracy of the prediction model.
By obtaining the time series data of the detection index, the prediction sequence data is obtained using a pre-trained time series prediction model based on the reverse sequence of the first time series data and the second time series data respectively. Then, the correlation between the first time period sequence data and the first time period prediction data, and the correlation between the second time period sequence data and the second time period prediction data, are determined, and whether the time series data is the target data of the target event is determined based on these correlations.
By determining the positive and reverse correlation between time series data and target events, data with strong correlation with target events can be effectively mined, improving the training effect and prediction accuracy of the sepsis prediction model.
Smart Images
Figure CN119905271A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of medical data mining, and specifically provides a data processing method, a training method for a sepsis prediction model, and an intelligent device. Background Art
[0002] In the prior art, prediction models based on time series data are applied to prediction tasks of various target events. For example, the target event may be the onset of sepsis, and the prediction model may be a sepsis prediction model, which may collect time series data obtained by various detection indicators of the observed object as input data to predict the time or probability of sepsis onset. Accordingly, when training such prediction models based on time series, it is necessary to collect multiple groups of time series data as training data, and the selection of training data will affect the training effect and prediction accuracy of the prediction model.
[0003] In the related technology, the selection of training data mainly considers the positive correlation between time series data and target events, so as to predict the future trend of target events through historical data. However, this method easily introduces training data with weak correlation with target events, which affects the training effect of the prediction model and reduces the accuracy of the prediction task. Summary of the invention
[0004] The present application aims to solve the above technical problem, that is, to solve the problem in the prior art of lack of mining of data that is highly correlated with the target event.
[0005] In a first aspect, the present application provides a data processing method, which comprises:
[0006] Acquire time series data of the detection index, the time series data including first time period series data and second time period series data arranged in time order; any one of the first time period series data and the second time period series data includes a detection index value corresponding to a characteristic time point of a target event;
[0007] Based on the reverse sequence of the first time period series data and the second time period series data, respectively, using a pre-trained time series prediction model to obtain prediction sequence data, the prediction sequence data including the first time period prediction data and the second time period prediction data;
[0008] Determine a first correlation between the first time period series data and the first time period forecast data, and determine a second correlation between the second time period series data and the second time period forecast data;
[0009] Based on the first correlation and the second correlation, it is determined whether the time series data is target data of the target event.
[0010] In some embodiments, the step of obtaining time series data of the detection index includes:
[0011] Acquire initial time series data of the detection index, wherein the initial time series data includes a plurality of detection index values, respective timestamps of the plurality of detection index values, and a characteristic time point label of the target event;
[0012] The initial time series data is segmented based on the characteristic time point labels to obtain the first time period series data and the second time period series data.
[0013] In some embodiments, the predicted sequence data is obtained by using a pre-trained time series prediction model based on the reverse sequence of the first time period series data and the second time period series data, and the predicted sequence data includes the first time period prediction data and the second time period prediction data, including:
[0014] Inputting the first time period series data into the pre-trained time series prediction model to obtain the second time period prediction data;
[0015] The reverse sequence of the second time period series data is input into the pre-trained time series prediction model to obtain the first time period prediction data.
[0016] In some embodiments, determining a first correlation between the first time period series data and the first time period prediction data includes:
[0017] Determine a first Pearson coefficient of the first time period series data and the first time period forecast data;
[0018] and / or,
[0019] Determining a second correlation between the second time period series data and the second time period forecast data includes:
[0020] A second Pearson coefficient of the second time period series data and the second time period prediction data is determined.
[0021] In some embodiments, determining whether the time series data is target data of the target event based on the first correlation and the second correlation includes:
[0022] Determine whether an average value of the sum of the absolute value of the first correlation and the absolute value of the second correlation is within a preset threshold range;
[0023] If yes, the time series data is determined as target data of the target event.
[0024] In some embodiments, the method further includes: when the average value of the sum of the absolute value of the first correlation and the absolute value of the second correlation is not within the preset threshold range, marking the time series data as data to be filtered.
[0025] In a second aspect, the present application provides a method for training a sepsis prediction model, comprising:
[0026] Acquire an initial training data set, wherein the initial training data set includes multiple training data subsets, and each of the multiple training data subsets includes time series data of at least one detection indicator;
[0027] For the time series data of the at least one detection indicator of the multiple training data subsets, respectively process the time series data using any of the data processing methods described above to determine whether it is target data for the onset of sepsis;
[0028] Screening out the time series data determined as the target data from the time series data corresponding to the multiple training data subsets, and constructing a target training data set based on the time series data determined as the target data;
[0029] The sepsis prediction model is trained based on the target training data set to obtain a trained sepsis prediction model.
[0030] In a third aspect, the present application provides a method for predicting sepsis, comprising:
[0031] Obtaining time series data of at least one detection indicator of the target observation object;
[0032] The time series data of the at least one detection indicator is input into a trained sepsis prediction model to obtain a prediction result; wherein the trained sepsis prediction model is trained based on the training method of the sepsis prediction model described above.
[0033] In a fourth aspect, the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, any of the methods described above is implemented.
[0034] In a fifth aspect, the present application provides a smart device, comprising:
[0035] at least one processor;
[0036] and, a memory communicatively coupled to the at least one processor;
[0037] The memory stores a computer program, and when the computer program is executed by the at least one processor, any of the above-mentioned methods is implemented.
[0038] In the case of adopting the above-mentioned technical scheme, the present application can obtain time series data of detection indicators, the time series data includes first time period series data and second time period series data arranged in chronological order; any one of the first time period series data and the second time period series data includes a detection indicator value corresponding to a characteristic time point of a target event; based on the reverse sequence of the first time period series data and the second time period series data, respectively, a pre-trained time series prediction model is used to obtain prediction sequence data, the prediction sequence data includes first time period prediction data and second time period prediction data; determine a first correlation between the first time period series data and the first time period prediction data, and determine a second correlation between the second time period series data and the second time period prediction data; based on the first correlation and the second correlation, determine whether the time series data is the target data of the target event. The method determines the positive correlation between the time series data and the target event by determining the second correlation of the time series data of the detection indicator. At the same time, the method determines the reverse correlation between the time series data and the target event by determining the first correlation of the time series data of the detection indicator. Through the positive correlation and the reverse correlation, it can be determined whether the current time series data and the target event are strongly correlated, that is, whether it is the target data of the target event, which is conducive to mining the time series data with strong correlation with the target event, and then can be used as training data for the prediction task of the target event, which is conducive to subsequently improving the training effect and prediction accuracy of the target event-related prediction model. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] The preferred embodiments of the present application are described below in conjunction with the accompanying drawings, in which:
[0040] Figure 1 It is a flowchart of a data processing method provided in an embodiment of the present application;
[0041] Figure 2 is a flowchart of a specific implementation method of step S11 provided in an embodiment of the present application;
[0042] Figure 3 is a flowchart of a specific implementation method of step S14 provided in an embodiment of the present application;
[0043] Figure 4 It is a flow chart of a training method for a sepsis prediction model provided in an embodiment of the present application;
[0044] Figure 5 is a flow chart of a sepsis prediction method provided in an embodiment of the present application;
[0045] Figure 6 It is a schematic diagram of the structure of the smart device provided in the embodiment of the present application. DETAILED DESCRIPTION
[0046] Some embodiments of the present application are described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principles of the present application and are not intended to limit the protection scope of the present application.
[0047] Based on the description in the background technology section, it can be seen that the selection of training data in the related technology mainly considers the positive correlation between time series data and target events, so as to predict the future trend of target events through historical data. However, this method easily introduces training data with weak correlation with the target event, which affects the training effect of the prediction model and reduces the accuracy of the prediction task.
[0048] In view of this, the present application provides a data processing method, which determines the positive correlation between the time series data and the target event by determining the second correlation of the time series data of the detection indicator. At the same time, it determines the reverse correlation between the time series data and the target event by determining the first correlation of the time series data of the detection indicator, so as to measure the correlation between the future trend of the target event and the historical data through the reverse correlation; at the same time, it determines whether the current time series data and the target event are strongly correlated, that is, whether it is the target data of the target event through the positive correlation and the reverse correlation. This method is conducive to mining time series data with a strong correlation with the target event, and then the time series data determined as the target data can be used as training data, which is conducive to subsequently improving the training effect and prediction accuracy of the target event related prediction model.
[0049] See also Figure 1 As shown, Figure 1 It is a flowchart of a data processing method provided by an embodiment of the present application, which may include:
[0050] Step S11: Acquire time series data of the detection index, the time series data including first time period series data and second time period series data arranged in chronological order;
[0051] Wherein, any one of the first time period series data and the second time period series data includes a detection index value corresponding to a characteristic time point of the target event;
[0052] Step S12: Based on the reverse sequence of the first time period series data and the second time period series data, respectively, a pre-trained time series prediction model is used to obtain prediction series data, where the prediction series data includes the first time period prediction data and the second time period prediction data;
[0053] Step S13: determining a first correlation between the first period series data and the first period forecast data, and determining a second correlation between the second period series data and the second period forecast data;
[0054] Step S14: Based on the first correlation and the second correlation, determine whether the time series data is target data of the target event.
[0055] In the embodiments of the present application, the detection index may be a detection index that can be collected in the scenario where the target event occurs. In some embodiments, the target event may be the onset of sepsis, and the detection index may be at least one index that can be detected after the observed object enters the ICU (Intensive Care Unit). As an example, the detection index may include any one of temperature, heart rate, and blood pressure.
[0056] In an embodiment of the present application, characteristic time points of target events can be pre-marked in time series data, and the characteristic time points of target events can be set according to actual application scenarios. In some embodiments, the characteristic time points can be used to characterize the time points at which target events occur or change. Taking the scenario of sepsis onset as an example, the characteristic time points can be the time points at which sepsis occurs or the time points at which sepsis changes.
[0057] In some embodiments, see Figure 2 As shown, step S11 may specifically be:
[0058] Step S111: Acquire initial time series data of the detection index, where the initial time series data includes multiple detection index values, respective timestamps of the multiple detection index values, and characteristic time point labels of target events;
[0059] Step S112: Segment the initial time series data based on the characteristic time point labels to obtain first time period series data and second time period series data.
[0060] Among them, the time series data of the detection index can be expressed as X = {x t1 、x t2 、x t3 、x t4 ……、x tn ,……x tm}, where x tm represents the detection index value corresponding to time point m, the characteristic time point can be tn, and the first time series data can be expressed as X 1 ={x t1 、x t2 、x t3 、x t4 ……x tn}, the second time series data can be expressed as X 2 ={x t(n+1) 、x t(n+2) 、x t(n+3) 、x t(n+4)……x tm}, m>n, m and n are both positive integers.
[0061] Both the first time period series data and the second time period series data may include detection index values of multiple time points, and the first time period series data and the second time period series data may contain detection index values of the same number of time points, or contain detection index values of different numbers of time points, which can be flexibly set according to the actual application scenario and the allowable input number of the time series prediction model.
[0062] In the embodiment of the present application, the time series prediction model can be pre-trained by the following steps:
[0063] For multiple observation objects where target events occur, the observed time series data of at least one detection indicator are obtained respectively to obtain a training data set; the time series prediction model is trained with the training data set as input to obtain a trained time series prediction model. The training of the time series prediction model based on the training data set can be specifically that the time series prediction model can randomly mask the observed time series data of the detection indicator with the training data set as input, and use the unmasked part of the observed time series data as training data to predict the masked part of the time series data, and learn the trend of the time series data of the detection indicator related to the target event through a self-supervised learning method.
[0064] In some embodiments, the time series prediction model can adopt the Timer model, which is a large time series model based on the Transformer decoder structure, which can be generatively pre-trained based on the observed time series data to obtain a time series analysis model that is universal for the task. In other embodiments, the time series model provided in this application can also be implemented using other large time series models.
[0065] In some embodiments, step S12 may specifically include:
[0066] Input the first period series data into a pre-trained time series prediction model to obtain the second period prediction data;
[0067] The reverse sequence of the second period series data is input into the pre-trained time series prediction model to obtain the prediction data for the first period.
[0068] Taking the second period series data shown above as an example, the reverse sequence of the second period series data can be expressed as {x tm ……x t(n+4) 、x t(n+3) 、x t(n+2) 、x t(n+1)}.
[0069] It should be noted that the above is only an example of first obtaining the forecast data for the second time period and then obtaining the forecast data for the first time period. In other embodiments, the forecast data for the first time period may be obtained first and then the forecast data for the second time period.
[0070] In some embodiments, determining the first correlation between the first time period series data and the first time period prediction data in step S13 may include: determining a first Pearson coefficient between the first time period series data and the first time period prediction data.
[0071] In some embodiments, determining a second correlation between the second period series data and the second period prediction data may include: determining a second Pearson coefficient between the second period series data and the second period prediction data.
[0072] The Pearson coefficient can be used to characterize the degree of correlation between two trends. When the Pearson coefficient is 1, it means that the two trends are synchronized. When the Pearson coefficient is -1, it means that the two trends are asynchronous. When the Pearson coefficient is 0, it means that there is no correlation between the two trends.
[0073] In the embodiment of the present application, by determining the second correlation, the positive correlation between the first time period series data and the second time period series data can be obtained, which can also be used to represent the positive correlation between the time series data of the detection indicator and the target event. When the second correlation is characterized by the second Pearson coefficient, the closer the second Pearson coefficient is to 1, the closer the trend of the second time period prediction data and the second time period series data is to synchronization, which can correspondingly indicate that the first time period series data and the second time period series data have a high positive correlation, and can predict more accurate second time period series data based on the first time period series data.
[0074] By determining the first correlation, the reverse correlation between the first time period series data and the second time period series data can be obtained, which can also be used to represent the reverse correlation between the time series data of the detection indicator and the target event. When the first correlation is characterized by the first Pearson coefficient, the closer the first Pearson coefficient is to 1, the closer the trend of the first time period prediction data and the first time period series data is to synchronization, which can correspondingly indicate that the reverse correlation between the first time period series data and the second time period series data is high, and more accurate first time period series data can be predicted based on the second time period series data.
[0075] In other embodiments, other methods for measuring the correlation between two trends may also be used to determine the first correlation and / or the second correlation.
[0076] In some embodiments, step S14 may specifically be to determine whether the time series data is target data of the target event according to whether the first correlation and the second correlation meet a preset condition.
[0077] In the embodiment of the present application, by combining the time series data of the detection index and the positive correlation and reverse correlation of the target event, it is determined whether the current time series data is data with a strong correlation with the target event, that is, the target data. When both the positive correlation and the reverse correlation are relatively high, it can be determined that the correlation is strong. Among them, the preset conditions can be flexibly set based on this and the application scenario.
[0078] In some embodiments, preset thresholds may be set for the first correlation and the second correlation respectively to determine whether the first correlation and the second correlation are greater than their respective corresponding preset thresholds. If so, they are determined to be target data.
[0079] In other embodiments, see Figure 3 As shown, step S14 may specifically be:
[0080] Step S141: determining whether the average value of the absolute value of the first correlation and the sum of the absolute values of the second correlation is within a preset threshold range;
[0081] If yes, execute step S142: determine the time series data as target data of the target event.
[0082] When the time series data is determined to be the target data of the target event, it is determined that the time series data and the target event are highly correlated, so that the mining of data with strong correlation to the target event can be achieved, and it can be used as training data for the subsequent training process of the prediction task for the target event.
[0083] When the first Pearson coefficient is used to represent the first correlation and the second Pearson coefficient is used to represent the second correlation, the average of the sum of the absolute value of the first correlation and the absolute value of the second correlation can be used to represent the correlation between the time series data and the target event: Among them, a represents the correlation between the time series data and the target event, p f represents the second Pearson coefficient, p b Represents the first Pearson coefficient.
[0084] In the embodiment of the present application, the purpose is to determine the target data with high correlation with the target event. When a represents the correlation between the time series data and the target event, the closer a is to 1, the higher the correlation. On this basis, the range of the preset threshold can be flexibly set according to the needs. As an example, the preset threshold range can be (0.5, 1].
[0085] In other embodiments, when the average of the sum of the absolute value of the first correlation and the absolute value of the second correlation is not within a preset threshold range, the method may further include: marking the time series data as data to be filtered to facilitate subsequent rapid data screening.
[0086] In some embodiments, after step S14, the method may further include: marking the time series data of the current detection index. When the time series data is target data, the time series data may be marked as a label of the target data to facilitate subsequent rapid data screening.
[0087] The above is a data processing method provided in an embodiment of the present application, which can obtain time series data of detection indicators, the time series data includes first time period series data and second time period series data arranged in chronological order; any one of the first time period series data and the second time period series data includes a detection indicator value corresponding to a characteristic time point of a target event; based on the reverse sequence of the first time period series data and the second time period series data, respectively, a pre-trained time series prediction model is used to obtain prediction sequence data, the prediction sequence data includes first time period prediction data and second time period prediction data; determine a first correlation between the first time period series data and the first time period prediction data, and determine a second correlation between the second time period series data and the second time period prediction data; based on the first correlation and the second correlation, determine whether the time series data is the target data of the target event. The method determines the positive correlation between the time series data and the target event by determining the second correlation of the time series data of the detection indicator. At the same time, the method determines the reverse correlation between the time series data and the target event by determining the first correlation of the time series data of the detection indicator. Through the positive correlation and the reverse correlation, it can be determined whether the current time series data and the target event are strongly correlated, that is, whether it is the target data of the target event, which is conducive to mining the time series data with strong correlation with the target event, and then can be used as training data for the prediction task of the target event, which is conducive to subsequently improving the training effect and prediction accuracy of the target event-related prediction model.
[0088] Another aspect of the present application provides a method for training a sepsis prediction model, see Figure 4 As shown, Figure 4 : is a flow chart of a training method for a sepsis prediction model provided in an embodiment of the present application, which may include:
[0089] Step S41: obtaining an initial training data set, the initial training data set including multiple training data subsets, each of which includes time series data of at least one detection indicator;
[0090] Step S42: for the time series data of at least one detection indicator of the plurality of training data subsets, respectively process the time series data using the data processing method provided in the embodiment of the present application to determine whether it is target data of sepsis onset;
[0091] Step S43: Screening out time series data determined as target data from the time series data corresponding to the plurality of training data subsets, and constructing a target training data set based on the time series data determined as target data;
[0092] Step S44: training the sepsis prediction model based on the target training data set to obtain a trained sepsis prediction model.
[0093] In some embodiments, step S41 may specifically be to obtain ICU observation data of multiple observation objects, and use the ICU observation data of each observation object as a training data subset, and each training data subset may include time series data of at least one detection index. The detection index may include any one of body temperature, blood pressure, heart rate, etc.
[0094] Step S42 may specifically be that when sepsis onset is taken as the target event, the data processing method provided in the above embodiment of the present application is used to process the time series data of each detection indicator of each training data subset respectively to determine whether the time series data of the current detection indicator is the target data of sepsis onset.
[0095] By adopting the data processing method provided in the embodiment of the present application, time series data with a strong correlation with the onset of sepsis can be effectively mined.
[0096] In some embodiments, step S42 may further include marking the time series data determined as target data, and the time series data may be marked as target data. Accordingly, in step S43, the time series data determined as target data may be screened out from the time series data corresponding to the plurality of training data subsets based on the marking.
[0097] In some other embodiments, after step S42, the time series data determined not to be target data may be marked as data to be filtered. Accordingly, in step S43, the time series data corresponding to the plurality of training data subsets may be filtered based on the marking of the data to be filtered, and the time series data retained after filtering is the time series data determined to be target data.
[0098] In the embodiments of the present application, the sepsis prediction model can be used for the prediction task of the probability or time of sepsis onset. In some embodiments, the sepsis prediction model can be implemented by selecting a model that can realize the corresponding function in the art based on actual needs, for example, a deep neural network can be used.
[0099] In other embodiments, a training data set may be directly constructed based on the time series data determined as the target data and used for training the sepsis prediction model, omitting the screening steps in steps S41, S42, and S43 to save processes and improve training efficiency.
[0100] By constructing a target training data set based on the time series data determined as the target data for training the sepsis prediction model, it is beneficial to avoid the interference of training data with weak correlation, and can effectively improve the training effect of the sepsis prediction model and the accuracy of the prediction.
[0101] In another aspect of the present application, a method for predicting sepsis is provided. Figure 5 As shown, Figure 5 : is a flow chart of a sepsis prediction method provided in an embodiment of the present application, which may include:
[0102] Step S51: Acquire time series data of at least one detection index of the target observation object;
[0103] Step S52: inputting the time series data of at least one detection indicator into the trained sepsis prediction model to obtain a prediction result.
[0104] The trained sepsis prediction model is obtained by training based on the training method of the sepsis prediction model provided in the above-mentioned embodiment of the present application.
[0105] Among them, when the prediction task corresponding to the sepsis prediction model is to predict the onset time of sepsis, the prediction result can be the predicted onset time; when the prediction task corresponding to the sepsis prediction model is the onset summary of sepsis, the prediction result can be the predicted onset probability. In other embodiments, the sepsis prediction model can also be applied to other prediction tasks related to sepsis.
[0106] The sepsis prediction method provided in the embodiments of the present application can effectively improve the accuracy of prediction.
[0107] It is understood by those skilled in the art that all or part of the processes in the method for implementing the above-mentioned embodiments of the present application can also be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable storage medium may include: any entity or device, medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory, random access memory, electric carrier signal, telecommunication signal and software distribution medium, etc. that can carry the computer program code.
[0108] Another aspect of the present application further provides a computer-readable storage medium, in which a computer program is stored, and when the computer program is executed by a processor, the method in any of the above embodiments can be implemented. The computer-readable storage medium can be a storage device formed by various electronic devices. Optionally, the computer-readable storage medium in the embodiment of the present application is a non-temporary computer-readable storage medium.
[0109] In another aspect of the present application, a smart device is provided.
[0110] In an embodiment of an intelligent device according to the present application, the intelligent device may include at least one processor; and a memory connected to the at least one processor in communication; wherein a computer program is stored in the memory, and when the computer program is executed by the at least one processor, the method described in any of the above embodiments is implemented. The intelligent device described in the present application may include a robot and other devices. Figure 6 , Figure 6 FIG. 6 exemplarily shows that the memory 61 and the processor 62 are communicatively connected via a bus.
[0111] In some embodiments of the present application, the smart device may further include at least one sensor, which is used to sense information and obtain time series data of detection indicators of the observed object. The sensor is communicatively connected to any type of processor mentioned in the present application.
[0112] The personal information of the relevant observation objects that may be involved in the various embodiments of this application shall be processed in strict accordance with the requirements of laws and regulations, follow the principles of legality, legitimacy and necessity, and based on the reasonable purposes of business scenarios, to process the personal information that the observation objects actively provide during the use of products / services or that is generated due to the use of products / services, as well as the personal information obtained with the authorization of the observation objects.
[0113] This application attaches great importance to the security of the personal information of the observed subjects, and has taken reasonable and feasible security protection measures that meet industry standards to protect the information of the observed subjects and prevent the information of the observed subjects from being accessed, disclosed, used, modified, damaged or lost without authorization.
[0114] So far, the technical solutions of the present application have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present application is obviously not limited to these specific embodiments. Without departing from the principles of the present application, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present application.
Claims
1. A data processing method, characterized in that: include: Acquire time series data of the detection index, wherein the time series data includes first time period series data and second time period series data arranged in chronological order; Any one of the first time period series data and the second time period series data includes a detection index value corresponding to a characteristic time point of a target event; Based on the reverse sequence of the first time period series data and the second time period series data, respectively, using a pre-trained time series prediction model to obtain prediction sequence data, the prediction sequence data including the first time period prediction data and the second time period prediction data; Determine a first correlation between the first time period series data and the first time period forecast data, and determine a second correlation between the second time period series data and the second time period forecast data; Based on the first correlation and the second correlation, it is determined whether the time series data is target data of the target event.
2. The method according to claim 1, characterized in that: The obtaining of time series data of the detection index includes: Acquire initial time series data of the detection index, wherein the initial time series data includes a plurality of detection index values, respective timestamps of the plurality of detection index values, and a characteristic time point label of the target event; The initial time series data is segmented based on the characteristic time point labels to obtain the first time period series data and the second time period series data.
3. The method according to claim 1, characterized in that The predicted sequence data is obtained by using a pre-trained time series prediction model based on the reverse sequence of the first time period series data and the second time period series data, and the predicted sequence data includes the first time period prediction data and the second time period prediction data, including: Inputting the first time period series data into the pre-trained time series prediction model to obtain the second time period prediction data; The reverse sequence of the second time period series data is input into the pre-trained time series prediction model to obtain the first time period prediction data.
4. The method according to claim 1, characterized in that: Determining a first correlation between the first time period series data and the first time period forecast data includes: Determine a first Pearson coefficient of the first time period series data and the first time period forecast data; and / or, Determining a second correlation between the second time period series data and the second time period forecast data includes: A second Pearson coefficient of the second time period series data and the second time period prediction data is determined.
5. The method according to any one of claims 1 to 4, characterized in that Determining whether the time series data is target data of the target event based on the first correlation and the second correlation includes: Determine whether an average value of the sum of the absolute value of the first correlation and the absolute value of the second correlation is within a preset threshold range; If yes, the time series data is determined as target data of the target event.
6. The method according to claim 5, characterized in that The method further includes: when the average value of the sum of the absolute value of the first correlation and the absolute value of the second correlation is not within the preset threshold range, marking the time series data as data to be filtered.
7. A method for training a sepsis prediction model, characterized in that: include: Acquire an initial training data set, wherein the initial training data set includes multiple training data subsets, and each of the multiple training data subsets includes time series data of at least one detection indicator; For the time series data of the at least one detection indicator of the multiple training data subsets, the time series data are processed by the method described in any one of claims 1 to 6 to determine whether it is target data of sepsis onset; Screening out the time series data determined as the target data from the time series data corresponding to the multiple training data subsets, and constructing a target training data set based on the time series data determined as the target data; The sepsis prediction model is trained based on the target training data set to obtain a trained sepsis prediction model.
8. A method for predicting sepsis, characterized in that: include: Obtaining time series data of at least one detection indicator of the target observation object; The time series data of the at least one detection indicator is input into a trained sepsis prediction model to obtain a prediction result; wherein the trained sepsis prediction model is trained based on the method described in claim 7.
9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.
10. A smart device, characterized in that: include: at least one processor; and, a memory communicatively coupled to the at least one processor; The memory stores a computer program, and when the computer program is executed by the at least one processor, the method according to any one of claims 1 to 8 is implemented.