Processing method and system for predicting delivery time and delivery mode by implementing data analysis based on machine learning

Abstract

The invention discloses a processing method and system for predicting delivery time and delivery mode by implementing data analysis based on machine learning. In the initial inspection stage, heart rate, blood pressure and body movement data of the pregnant woman are monitored through the wearable device interface; past medical history information of the pregnant woman is extracted from the electronic health recording system; obtaining a development index of the fetus through ultrasonic equipment; performing preprocessing operation on the acquired pregnancy data to obtain a unified multi-source data set; extracting features having high correlation with the delivery time and mode from the data of the preprocessed unified multi-source data set, and recording the features as target features; training a machine learning model based on the selected target features to obtain a trained target prediction model; and inputting latest collected to-be-predicted pregnancy data, performing prediction processing operation on the delivery time and the delivery mode by using the trained target prediction model, and outputting the delivery time and the delivery mode. The system can comprehensively monitor and reflect the data of the unique health condition of each pregnant woman, achieves the precise prediction of the delivery time and the delivery mode through the cooperative work of all modules, and provides data technical support for the subsequent production.

Description

Technical Field

[0001] The present invention relates to the field of pregnant women's delivery monitoring, and in particular to a processing method and system for predicting delivery time and mode based on data analysis implemented by machine learning. Background Art

[0002] With the continuous improvement of medical standards and the enhancement of maternal health awareness, the importance of pregnancy health monitoring in ensuring the safety of mothers and babies has become increasingly prominent. Pregnancy is a special stage in a woman's life, involving the dual health of the mother and the fetus. Through scientific and systematic health monitoring, various health risks that may occur during pregnancy can be effectively prevented and managed to ensure the safety of mothers and babies. Therefore, pregnancy health monitoring is not only an important part of medical services, but also a key link in the public health system.

[0003] In the later stages of pregnancy, predicting the time and mode of delivery is of great significance to both pregnant women and medical institutions. Accurate delivery time prediction can help medical institutions rationally arrange obstetric resources, optimize hospital surgery schedules, and ensure timely and adequate medical support during delivery. At the same time, accurate prediction can also reduce the occurrence of emergency deliveries, reduce medical costs, and improve the efficiency and quality of medical services.

[0004] The researchers also found that predicting the mode of delivery is equally important. Natural childbirth and caesarean section each have their own indications and risks. Predicting the mode of delivery in advance can help doctors and pregnant women make adequate preparations and choose the most suitable mode of delivery to reduce possible complications during delivery and improve the safety of mother and baby. For example, for pregnant women who are inclined to have a caesarean section, making relevant preparations in advance can avoid emergency surgery and improve the success rate and safety of the operation.

[0005] Limitations of traditional delivery time prediction methods. Generally speaking, traditional delivery time prediction methods mainly rely on medical experience and routine examinations. These methods have certain limitations in practical applications, which are specifically manifested in the following aspects: First, traditional methods often rely on doctors' clinical experience and observation of pregnant women's physical signs for prediction. Although experienced doctors can improve the accuracy of predictions to a certain extent, this method is highly subjective and dependent. Different doctors may have different prediction results due to different experiences and judgment criteria, which affects the consistency and reliability of predictions to a certain extent.

[0006] Second: The prediction model of a single physiological parameter has limited accuracy. Some studies have attempted to use a single or a few physiological parameters (such as cervical length, body temperature, etc.) to establish statistical or machine learning models for delivery prediction. However, the health status of pregnancy and the delivery process are affected by many factors, and there are complex relationships and multi-factor interactions between these physiological parameters. Prediction models that rely on a single parameter often cannot fully reflect the health status of pregnancy, and their prediction accuracy and reliability are therefore limited.

[0007] Third: Lack of intelligent and personalized monitoring methods; existing delivery prediction methods mostly rely on fixed-point and regular medical examinations, and lack continuous, dynamic and personalized monitoring methods.

[0008] Therefore, most existing methods rely on fixed-point and regular medical examinations, lack continuous, dynamic and personalized monitoring methods, and it is difficult to capture the health changes of pregnant women and fetuses in real time. Summary of the invention

[0009] The purpose of the present invention is to provide a processing method and system for predicting the time and mode of delivery based on data analysis implemented by machine learning, which solves the above-mentioned technical problems pointed out in the prior art.

[0010] The present invention provides a processing method for predicting delivery time and delivery mode by implementing data analysis based on machine learning, comprising the following steps:

[0011] Data collection during pregnancy: monitor the heart rate, blood pressure and body movement data of pregnant women through wearable device interfaces; extract the medical history information of pregnant women from the electronic health record system; obtain fetal development indicators through ultrasound equipment;

[0012] Perform preprocessing operations on the collected pregnancy data to obtain a unified multi-source data set;

[0013] Extract features with high correlation with delivery time and mode from the preprocessed unified multi-source data set and record them as target features;

[0014] Use the machine learning model trained based on the selected target features to obtain a trained target prediction model;

[0015] Input the latest collected pregnancy data to be predicted, use the trained target prediction model to predict the delivery time and mode of delivery, and output the delivery time and mode.

[0016] Preferably, as an implementable embodiment, when the pregnancy data collection is executed, it also includes obtaining the physiological indicators of the pregnant woman through the wearable device interface; the physiological indicators of the pregnant woman include the heart rate, blood pressure and body movement data of the pregnant woman; the body movement data is the physical activity data recorded by the sensor device, including the number of steps walked per day and the length of sleep;

[0017] It also includes calling the database of the hospital's electronic health record system to obtain the medical history information of multiple pregnant women (and the medical history information of the pregnant woman to be tested is also obtained from the database of the electronic health record system); the medical history information of the pregnant woman includes chronic disease history information;

[0018] The method also includes calling a data interface of an ultrasonic device to obtain fetal development indicators; the fetal development indicators include fetal weight, head circumference, and abdominal circumference.

[0019] Preferably, as an implementable embodiment, the preprocessing operation performed on the collected pregnancy data to obtain a unified multi-source data set specifically includes:

[0020] The denoising processing unit performs filtering processing on the collected heart rate, blood pressure data and body movement data to obtain filtered sampling data;

[0021] The data integration unit aligns and integrates the filtered sampled data from the wearable device monitoring, the database of the hospital's electronic health record system, and the ultrasound imaging data according to the timestamp, and then outputs a unified multi-source data set;

[0022] Preferably, as an implementable scheme, the features with high correlation with the time and mode of delivery are extracted from the pre-processed unified multi-source data set and recorded as target features, specifically including:

[0023] Statistical feature calculation unit: input the preprocessed unified multi-source data set, and then calculate the basic statistical features of each physiological indicator, including the mean, standard deviation, maximum value and minimum value; calculate the basic statistical features of each fetal developmental indicator, including the mean, standard deviation, maximum value and minimum value; summarize the basic statistical features of the above physiological indicators and the basic statistical features of the developmental indicators to obtain a statistical feature data set;

[0024] Correlation analysis unit: input the statistical feature data set, use the Pearson correlation coefficient to evaluate the correlation between each basic statistical feature and the delivery time and mode of delivery, and screen out key features with a correlation coefficient higher than 0.7 as a high-correlation feature set;

[0025] Feature optimization unit: Then the high correlation feature set is input, and the recursive feature elimination (RFE) method is applied to further optimize the key features and finally select the N most predictive features as the target features.

[0026] Preferably, as an implementable method, a trained target prediction model is obtained by training a machine learning model based on the selected target features, specifically including:

[0027] The machine learning algorithm selection unit selects the random forest algorithm as the prediction model and records it as the random forest model;

[0028] Then the data set partitioning unit partitions the preprocessed data set into a training set and a test set in a ratio of 80:20;

[0029] The model validation and adjustment unit optimizes the key parameters of the random forest model through a 5-fold cross validation method; the key parameters include the number of trees (number of trees) and the maximum depth;

[0030] Feature weight calculation unit: Based on the formula group F = {f1, f2, ..., f 10}, calculate the weight of each feature, where

[0031] The model input unit provides the calculated feature weights as input parameters to the random forest model, ensuring that the random forest model receives optimized feature data;

[0032] The model optimization unit further adjusts the internal parameters of the random forest model to finally obtain a trained target prediction model; the internal parameters include sample weight allocation.

[0033] Preferably, as an implementable method, the latest collected pregnancy data to be predicted is input, and the delivery time and mode are predicted and processed using the trained target prediction model, and the delivery time and mode are output, which specifically includes:

[0034] The data input unit inputs the latest collected pregnancy data to be predicted, wherein the input latest collected pregnancy data to be predicted includes physiological indicators of pregnant women monitored in real time and fetal development indicators obtained by the latest ultrasonic equipment;

[0035] The prediction calculation unit uses the trained random forest model to calculate the input data and generate the prediction result Y=W·X+b;

[0036] Wherein Y represents the predicted delivery time and delivery mode; the delivery time is in days; the delivery mode includes natural delivery or caesarean section;

[0037] The result output unit transmits the prediction result to the user interface module, and displays the predicted time and mode of delivery to the pregnant woman and medical staff through the user interface module.

[0038] Preferably, as an implementable embodiment, the user interface module includes any one of a mobile application APP or an electronic health record system.

[0039] Preferably, as an implementable embodiment; the user interface module establishes a communication connection with the target prediction model in real time; the user interface module updates the prediction results sent by the target prediction model in real time; and generates detailed report information containing predicted delivery time and mode of delivery from the prediction results.

[0040] Preferably, as an implementable method, after inputting the latest collected pregnancy data to be predicted, using the trained target prediction model to perform prediction processing operations on the delivery time and delivery mode, and outputting the delivery time and mode, it also includes:

[0041] Monitor the personal information of the pregnant women to be tested corresponding to the latest pregnancy data to be predicted;

[0042] Obtaining the medical history information of the pregnant woman to be tested extracted from the electronic health record system corresponding to the personal information of the pregnant woman to be tested;

[0043] Determine whether the pregnant woman to be tested currently has a history of chronic diseases. If so, mark the pregnant woman to be tested currently as a target pregnant woman for key attention.

[0044] Accordingly, the present invention provides a processing system for predicting delivery time and mode of delivery by implementing data analysis based on machine learning, including a data acquisition module, a data preprocessing module, a feature extraction module, a model training module and a prediction module:

[0045] Data collection module, used to collect pregnancy data;

[0046] A data preprocessing module is used to perform preprocessing operations on the collected pregnancy data to obtain a unified multi-source data set;

[0047] A feature extraction module is used to extract features with a high correlation with the time and mode of delivery from the preprocessed unified multi-source data set and record them as target features;

[0048] A model training module is used to train a machine learning model based on selected target features to obtain a trained target prediction model;

[0049] The prediction module is used to input the latest collected pregnancy data to be predicted, use the trained target prediction model to perform prediction processing operations on the delivery time and mode of delivery, and output the delivery time and mode.

[0050] Compared with the prior art, the embodiments of the present invention have at least the following technical advantages:

[0051] From the analysis of the above-mentioned processing method and system for predicting delivery time and mode of delivery based on machine learning data analysis provided by the present invention, it can be known that in specific applications, pregnancy data collection: monitoring the heart rate, blood pressure and body movement data of pregnant women through the wearable device interface; extracting the medical history information of pregnant women from the electronic health record system; obtaining fetal development indicators through ultrasonic equipment; performing preprocessing operations on the collected pregnancy data to obtain a unified multi-source data set; extracting features with a high correlation with delivery time and mode from the data of the preprocessed unified multi-source data set and recording them as target features; using a machine learning model based on the selected target features to obtain a trained target prediction model; inputting the latest collected pregnancy data to be predicted, using the trained target prediction model to perform predictive processing operations on the delivery time and mode of delivery, and outputting the delivery time and mode.

[0052] Analysis of the above technical solutions shows that the existing delivery prediction methods are difficult to integrate multi-source and multi-dimensional pregnancy health data, resulting in the prediction model being unable to fully reflect the health status of pregnant women and fetuses. At the same time, there is a lack of real-time monitoring methods, and it is impossible to capture the dynamic changes of health indicators in a timely manner.

[0053] However, the present invention monitors the heart rate, blood pressure and body movement data of pregnant women in real time through the wearable device interface; extracts the medical history information of pregnant women from the electronic health record system; and obtains fetal development indicators through ultrasonic equipment. This diversified data collection method ensures the diversity and comprehensiveness of data sources, covering physiological data, medical history data and fetal development data, thereby building a comprehensive foundation for pregnancy health information. The application of wearable devices realizes real-time monitoring of pregnant women's physiological indicators and can continuously capture the dynamic changes in pregnancy health status. This overcomes the problem of insufficient timeliness of regular inspections in traditional methods and ensures the real-time and continuity of data.

[0054] Then, the collected multi-source data is preprocessed and integrated into a unified multi-source data set. Through data cleaning, format unification and time alignment, the inconsistency of different data sources in format and time scale is solved, and a structured high-quality data set that can be used for subsequent analysis is formed. Through the above steps, the present invention effectively integrates multi-source data, ensures the real-time and comprehensiveness of the data, and solves the problem of insufficient data integration and real-time performance in the prior art.

[0055] At the same time, traditional prediction methods and single parameter models have low prediction accuracy under complex pregnancy health conditions and are difficult to provide reliable prediction results, which affects clinical decision-making.

[0056] The present invention selects high-correlation features: the Pearson correlation coefficient is used to evaluate the correlation between each feature and the time and mode of delivery, and the key features with a correlation coefficient higher than 0.7 are screened out; the recursive feature elimination (RFE) method is applied to further optimize the feature set, and finally the 10 most predictive features are selected. This feature selection process ensures that the data input into the model has a high degree of correlation and predictive ability, and improves the overall performance of the model. The random forest algorithm is used as the main prediction model. Random forest has the ability to process high-dimensional and diverse data, can capture complex patterns and nonlinear relationships in the data, and has good anti-overfitting capabilities. The model parameters (such as the number of trees and the maximum depth) are optimized through 5-fold cross validation to ensure the stability and generalization ability of the model during the training process. This optimization process improves the accuracy and reliability of the model and ensures its consistent performance on different data sets.

[0057] Finally, the trained random forest model is used to predict the time and mode of delivery based on the latest pregnancy data input. The model generates more robust and accurate prediction results through the integration of multiple decision trees, significantly improving the accuracy and reliability of the prediction.

[0058] Through the optimization of feature extraction and the application of advanced machine learning algorithms, the present invention significantly improves the accuracy and reliability of the prediction model and solves the problem of insufficient prediction accuracy in traditional methods. At the same time, by integrating multi-source data (physiological indicators, medical history, fetal development indicators), the present invention can fully reflect the unique health status of each pregnant woman and provide a basis for the construction of a personalized model.

[0059] In summary, the present invention provides a processing method for predicting the time and mode of delivery through data analysis based on machine learning. The system integrates multi-source data, extracts personalized features, and performs accurate predictive analysis, thereby achieving full monitoring and utilization of pregnancy data of different pregnant women and production prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 A schematic diagram of the main process of a method for predicting delivery time and delivery mode by implementing data analysis based on machine learning;

[0061] Figure 2 A schematic diagram of a specific process of synchronization of data collection during pregnancy in a method for predicting delivery time and delivery mode through data analysis based on machine learning;

[0062] Figure 3 A schematic diagram of a specific pre-processing process in a method for predicting delivery time and mode of delivery through data analysis based on machine learning;

[0063] Figure 4A schematic diagram of the specific process of extracting features from a method for predicting delivery time and mode through data analysis based on machine learning;

[0064] Figure 5 A schematic diagram of a specific process of obtaining a trained target prediction model in a method for predicting delivery time and delivery mode through data analysis based on machine learning;

[0065] Figure 6 A schematic diagram of a specific process of outputting the time and mode of delivery in a method for predicting the time and mode of delivery through data analysis based on machine learning;

[0066] Figure 7 A schematic diagram of the principle of a processing system for predicting the time and mode of delivery through data analysis based on machine learning.

[0067] Reference numerals: data acquisition module 10 ; data preprocessing module 20 ; feature extraction module 30 ; model training module 40 ; prediction module 50 . DETAILED DESCRIPTION

[0068] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0069] The present invention is further described in detail below through specific embodiments in conjunction with the accompanying drawings.

[0070] Embodiment 1

[0071] like Figure 1 As shown, the first embodiment of the present invention provides a processing method for predicting delivery time and delivery mode based on data analysis implemented by machine learning, including the following steps:

[0072] S101. Data collection during pregnancy: monitor the heart rate, blood pressure and body movement data of pregnant women through wearable device interfaces; extract the medical history information of pregnant women from the electronic health record system; obtain fetal development indicators through ultrasound equipment;

[0073] S102, performing preprocessing operations on the collected pregnancy data to obtain a unified multi-source data set;

[0074] S103, extracting features with a high correlation with delivery time and mode from the preprocessed unified multi-source data set and recording them as target features;

[0075] S104, using a machine learning model trained based on the selected target features to obtain a trained target prediction model;

[0076] S105, input the latest collected pregnancy data to be predicted, use the trained target prediction model to perform prediction processing operations on the delivery time and mode of delivery, and output the delivery time and mode.

[0077] Embodiment 1 of the present invention provides a processing method for predicting delivery time and mode of delivery by implementing data analysis based on machine learning, and the main method flow includes: Data collection (S101): Collect multi-source data through smart devices and hospital systems. Data preprocessing (S102): Clean, integrate and reduce the dimension of data. Feature extraction (S103): Calculate statistical features, screen and optimize key features. Model training (S104): Select random forest algorithm, optimize and train the model. Prediction processing (S105): Input the latest data, generate and output the prediction results of delivery time and mode.

[0078] Analysis of the above technical solutions shows that the existing delivery prediction methods are difficult to integrate multi-source and multi-dimensional pregnancy health data, resulting in the prediction model being unable to fully reflect the health status of pregnant women and fetuses. At the same time, there is a lack of real-time monitoring methods, and it is impossible to capture the dynamic changes of health indicators in a timely manner.

[0079] However, the present invention monitors the heart rate, blood pressure and body movement data of pregnant women in real time through the wearable device interface; extracts the medical history information of pregnant women from the electronic health record system; and obtains fetal development indicators through ultrasonic equipment. This diversified data collection method ensures the diversity and comprehensiveness of data sources, covering physiological data, medical history data and fetal development data, thereby building a comprehensive foundation for pregnancy health information. The application of wearable devices realizes real-time monitoring of pregnant women's physiological indicators and can continuously capture the dynamic changes in pregnancy health status. This overcomes the problem of insufficient timeliness of regular inspections in traditional methods and ensures the real-time and continuity of data.

[0080] Then, the collected multi-source data is preprocessed and integrated into a unified multi-source data set. Through data cleaning, format unification and time alignment, the inconsistency of different data sources in format and time scale is solved, and a structured high-quality data set that can be used for subsequent analysis is formed. Through the above steps, the present invention effectively integrates multi-source data, ensures the real-time and comprehensiveness of the data, and solves the problem of insufficient data integration and real-time performance in the prior art.

[0081] At the same time, researchers found that traditional prediction methods and single-parameter models have low prediction accuracy under complex pregnancy health conditions and are unable to provide reliable prediction results, which affects clinical decision-making.

[0082] The present invention selects high-correlation features: the Pearson correlation coefficient is used to evaluate the correlation between each feature and the time and mode of delivery, and the key features with a correlation coefficient higher than 0.7 are screened out; the recursive feature elimination (RFE) method is applied to further optimize the feature set, and finally the 10 most predictive features are selected. This feature selection process ensures that the data input into the model has a high degree of correlation and predictive ability, and improves the overall performance of the model. The random forest algorithm is used as the main prediction model. Random forest has the ability to process high-dimensional and diverse data, can capture complex patterns and nonlinear relationships in the data, and has good anti-overfitting capabilities. The model parameters (such as the number of trees and the maximum depth) are optimized through 5-fold cross validation to ensure the stability and generalization ability of the model during the training process. This optimization process improves the accuracy and reliability of the model and ensures its consistent performance on different data sets.

[0083] Finally, the trained random forest model is used to predict the time and mode of delivery based on the latest pregnancy data input. The model generates more robust and accurate prediction results through the integration of multiple decision trees, significantly improving the accuracy and reliability of the prediction.

[0084] Through the optimization of feature extraction and the application of advanced machine learning algorithms, the present invention significantly improves the accuracy and reliability of the prediction model and solves the problem of insufficient prediction accuracy in traditional methods. At the same time, by integrating multi-source data (physiological indicators, medical history, fetal development indicators), the present invention can fully reflect the unique health status of each pregnant woman and provide a basis for the construction of a personalized model.

[0085] In summary, the present invention provides a processing method and system for predicting the time and mode of delivery through data analysis based on machine learning, which realizes full monitoring and utilization of pregnancy data of different pregnant women and production prediction through the integration of multi-source data, extraction of personalized features and accurate predictive analysis.

[0086] like Figure 2 As shown, when the pregnancy data collection is executed, it also includes:

[0087] S1011. Acquire physiological indicators of the pregnant woman through a wearable device interface; the physiological indicators of the pregnant woman include heart rate, blood pressure and body movement data of the pregnant woman; the body movement data is physical activity data recorded by a sensor device, including the number of steps walked per day and the length of sleep;

[0088] S1012, further comprising calling the database of the electronic health record system of the hospital to obtain medical history information of multiple pregnant women (and the medical history information of the current pregnant woman to be tested is also obtained from the database of the electronic health record system); the medical history information of the pregnant woman includes chronic disease history information;

[0089] S1013, further comprising calling a data interface of an ultrasonic device to obtain fetal development indicators; the fetal development indicators include fetal weight, head circumference, and abdominal circumference.

[0090] It should be noted that in the above specific technical solutions, wearable device interface: use smart bracelets to monitor the heart rate, blood pressure and body movement data of pregnant women to ensure the continuity and accuracy of the data. Database connection of the hospital's electronic health record system: extract the medical history information of pregnant women from the hospital's electronic health record system, including but not limited to chronic diseases such as diabetes and hypertension. Ultrasonic imaging data interface: obtain fetal development indicators such as fetal weight, head circumference, abdominal circumference, etc. through ultrasonic equipment to evaluate the health status and development progress of the fetus.

[0091] Specifically, the pregnancy health monitoring of the embodiment of the present invention covers the dynamic monitoring of various physiological indicators of pregnant women, including heart rate, blood pressure, body temperature, body movement, etc., and also involves the assessment of fetal development, such as fetal weight, head circumference, abdominal circumference, etc. The real-time acquisition and analysis of these data not only helps to timely discover and deal with potential health problems, but also can provide personalized health management suggestions for pregnant women, optimize their daily living habits, and improve their overall health level.

[0092] Traditional prediction methods and single parameter models have low prediction accuracy under complex pregnancy health conditions and are difficult to provide reliable prediction results, which affects clinical decision-making; therefore, the technical solution adopted in the embodiment of the present invention collects a large amount of pregnancy data.

[0093] like Figure 3 As shown, the preprocessing operation performed on the collected pregnancy data to obtain a unified multi-source data set specifically includes:

[0094] S1121, the denoising processing unit performs filtering processing on the collected heart rate and blood pressure data and body movement data to obtain filtered sampling data;

[0095] S1122. The data integration unit aligns and integrates the filtered sampled data from the wearable device monitoring, the database of the hospital's electronic health record system, and the ultrasound imaging data according to the timestamps, and then outputs a unified multi-source data set.

[0096] It should be noted that in the above specific technical solution, the denoising unit filters the collected heart rate and blood pressure data to remove environmental noise and equipment interference to ensure the clarity and reliability of the data. The data integration unit aligns the data from the wearable device, the database of the hospital's electronic health record system, and the ultrasound imaging data according to the timestamp and integrates them into a unified multi-source data set.

[0097] like Figure 4As shown in the figure, the features with high correlation with delivery time and mode are extracted from the preprocessed unified multi-source data set and recorded as target features, including:

[0098] S1031, statistical feature calculation unit: input the preprocessed unified multi-source data set, and then calculate the basic statistical features of each physiological indicator, including the mean, standard deviation, maximum value and minimum value; calculate the basic statistical features of each fetal development indicator, including the mean, standard deviation, maximum value and minimum value; summarize the basic statistical features of the above physiological indicators and the basic statistical features of the development indicators to obtain a statistical feature data set;

[0099] S1032, correlation analysis unit: input the statistical feature data set, use the Pearson correlation coefficient to evaluate the correlation between each basic statistical feature and the delivery time and mode of delivery, and screen out key features with a correlation coefficient higher than 0.7 as a high correlation feature set;

[0100] S1033, feature optimization unit: then input the high correlation feature set, apply the recursive feature elimination (RFE) method, further optimize the key features, and finally select the N most predictive features as target features.

[0101] It should be noted that, in the above specific technical solution, the input of the statistical feature calculation unit: the preprocessed unified multi-source data set. The output of the statistical feature calculation unit: the basic statistical features of each physiological indicator (mean, standard deviation, maximum value, minimum value). The basic statistical features of the developmental indicators of each fetus are calculated, including the mean, standard deviation, maximum value and minimum value; the basic statistical features of the above physiological indicators and the basic statistical features of the developmental indicators are summarized to obtain a statistical feature data set;

[0102] Correlation analysis: Input: Statistical feature data set. Output: Filter out key features with correlation > 0.7 with delivery time and mode.

[0103] Feature optimization: Input: Highly relevant feature set. Output: The N most predictive features after RFE optimization.

[0104] It should be noted that in the above specific technical solution, the correlation analysis unit: uses the Pearson correlation coefficient to evaluate the correlation between each feature and the delivery time and mode of delivery, and screens out key features with a correlation coefficient higher than 0.7. The feature optimization unit: applies the recursive feature elimination (RFE) method to further optimize the feature set, and finally selects the N most predictive features to improve the performance of the model.

[0105] For example, in the "Delivery Time and Mode Prediction Method Based on Random Forest Algorithm", the feature extraction module is responsible for extracting features with high correlation with delivery time and mode from the preprocessed unified multi-source data set. This process is divided into three key steps: statistical feature calculation unit, correlation analysis unit and feature optimization unit. The logic of each step and its relationship will be explained in detail below.

[0106] Step 1: Statistical feature calculation unit; Objective: Extract basic statistical features from each physiological indicator to quantify the distribution characteristics of the data and provide a basis for subsequent correlation analysis and feature selection.

[0107] Specific operations:

[0108] Select physiological indicators:

[0109] For example, heart rate, blood pressure, body temperature, number of steps in body movement data, etc.

[0110] Compute statistical features:

[0111] Mean: reflects the central trend of the data.

[0112] Standard Deviation: It measures the dispersion of data, that is, the deviation of data points from the mean.

[0113] Maximum: The highest value in the data, which can reflect abnormally high values.

[0114] Minimum: The lowest value in the data, which can reflect abnormally low values.

[0115] Example:

[0116] Suppose the heart rate data of a pregnant woman after preprocessing is: [80,85,78,90,88].

[0117] Mean = (80 + 85 + 78 + 90 + 88) / 5 = 84.2;

[0118] Standard deviation ≈5.98;

[0119] Maximum value = 90;

[0120] Minimum value = 78;

[0121] Output:

[0122] A set of statistical features is generated for each physiological indicator, such as:

[0123] Mean heart rate: 84.2

[0124] Heart rate standard deviation: 5.98;

[0125] Maximum heart rate: 90;

[0126] Heart rate minimum: 78;

[0127] Step 2: Correlation analysis unit;

[0128] Objective: Evaluate the correlation between each statistical feature and the target variable (delivery time and mode of delivery), screen out features with high correlation with the target variable, and provide a basis for subsequent feature optimization.

[0129] Specific operations:

[0130] Define the target variable:

[0131] Delivery Time: A numeric variable with days as the unit.

[0132] Delivery Mode: Categorical variables, such as vaginal delivery or cesarean section.

[0133] Calculate the correlation:

[0134] For numerical target variables (delivery time): use the Pearson Correlation Coefficient to evaluate the linear correlation between each feature and delivery time. For categorical target variables (delivery mode): the Point-Biserial Correlation Coefficient or other correlation indicators suitable for categorical variables can be used.

[0135] Filter high correlation features:

[0136] A threshold value (such as 0.7) is set to filter out features whose absolute value of the correlation coefficient is greater than 0.7.

[0137] Example:

[0138] Suppose that after calculation, we find that:

[0139] The Pearson correlation coefficient between the mean heart rate and delivery time is 0.75 → the screening condition is met.

[0140] The Pearson correlation coefficient between the standard deviation of blood pressure and delivery time was 0.65 → the screening condition was not met.

[0141] Output:

[0142] A series of statistical features with a correlation greater than 0.7 with the time and mode of delivery were screened out, such as:

[0143] mean heart rate (correlation coefficient 0.75);

[0144] mean blood pressure (correlation coefficient 0.72);

[0145] ...(other qualifying features);

[0146] Step 3: Feature optimization unit;

[0147] Goal: Further optimize the feature set, remove redundant or unnecessary features, and ultimately select the most predictive features to improve model performance.

[0148] Specific operations:

[0149] Apply Recursive Feature Elimination (RFE): RFE is a feature selection method that recursively trains the model and removes the least important features until a predetermined number of features is reached.

[0150] process:

[0151] Train a random forest model using all candidate features.

[0152] Assess the importance of each feature.

[0153] Remove the least important features.

[0154] The above steps are repeated until a predetermined number of features (10 in this example, ie, N=10) remain.

[0155] Select the 10 most predictive features:

[0156] Through multiple rounds of iterations, RFE retains the features that contribute most to model prediction, ensuring the efficiency and relevance of the final feature set.

[0157] Example:

[0158] Assume that after correlation analysis, 15 features satisfy correlation > 0.7. After applying RFE, the following 10 features are finally selected:

[0159] Heart rate mean;

[0160] Mean blood pressure;

[0161] Mean fetal weight;

[0162] Mean fetal head circumference;

[0163] Mean number of steps in body motion data

[0164] ...(5 other features)

[0165] Output:

[0166] The final 10 most predictive features were identified, providing high-quality input to the random forest model.

[0167] like Figure 5 As shown, a trained target prediction model is obtained by training a machine learning model based on the selected target features, including:

[0168] S1041, the machine learning algorithm selection unit selects a random forest algorithm as a prediction model and records it as a random forest model;

[0169] S1042, then the data set division unit divides the preprocessed data set into a training set and a test set in a ratio of 80:20;

[0170] S1043, the model validation and adjustment unit optimizes the key parameters of the random forest model through a 5-fold cross validation method; the key parameters include the number of trees (number of trees) and the maximum depth;

[0171] S1044, feature weight calculation unit: based on the formula group F = {f1, f2, ..., f 10}, calculate the weight of each feature, where

[0172] S1045. The model input unit provides the calculated feature weights as input parameters to the random forest model, ensuring that the random forest model receives optimized feature data;

[0173] S1046. The model optimization unit further adjusts the internal parameters of the random forest model to finally obtain a trained target prediction model; the internal parameters include sample weight distribution.

[0174] It should be noted that in the above specific technical solutions, the machine learning algorithm selection unit: selects the random forest algorithm as the main prediction model because it has good performance in processing high-dimensional data and complex relationships. The data set division unit: divides the preprocessed data set into a training set and a test set in a ratio of 80:20 to ensure the effectiveness of model training and evaluation.

[0175] Model validation and adjustment unit: Through the 5-fold cross-validation method, the key parameters of the random forest model, such as the number of trees (number of trees) and the maximum depth, are optimized to improve the generalization ability of the model. Feature weight calculation unit: Based on the formula group F = {f1, f2, ..., f 10}, calculate the weight of each feature, where To quantify the importance of each feature.

[0176] Model input unit: Provides the calculated feature weights as input parameters to the random forest model to ensure that the model receives optimized feature data. Model optimization unit: Further adjusts the internal parameters of the random forest model, such as node splitting criteria and sample weight allocation, to minimize prediction errors and improve model accuracy.

[0177] like Figure 6 As shown, the latest collected pregnancy data to be predicted is input, and the trained target prediction model is used to predict the delivery time and mode of delivery, and the delivery time and mode are output, including:

[0178] S1051, the data input unit inputs the latest collected pregnancy data to be predicted, wherein the input latest collected pregnancy data to be predicted includes physiological indicators of pregnant women monitored in real time and fetal development indicators obtained by the latest ultrasound equipment;

[0179] S1052, the prediction calculation unit uses the trained random forest model to calculate the input data and generate a prediction result Y=W·X+b;

[0180] Wherein Y represents the predicted delivery time and delivery mode; the delivery time is in days; the delivery mode includes natural delivery or caesarean section;

[0181] S1053. The result output unit transmits the prediction result to the user interface module, and displays the predicted time and mode of delivery to the pregnant woman and medical staff through the user interface module.

[0182] It should be noted that, in the above specific technical solution, step S105: prediction processing operation; data input unit: input the latest collected pregnancy data, including real-time monitored physiological indicators (such as heart rate, blood pressure) and the latest ultrasonic imaging data (such as fetal development). Prediction calculation unit: use the trained random forest model to calculate the input data to generate a prediction result Y=W·X+b, where Y represents the predicted delivery time (in days) and delivery method (natural delivery or cesarean section). Result output unit: transmit the prediction result to the user interface module, and display the predicted delivery time and method to pregnant women and medical staff through the user interface module or other display methods.

[0183] Preferably, as an implementable embodiment, the user interface module includes any one of a mobile application APP or an electronic health record system.

[0184] Preferably, as an implementable embodiment; the user interface module establishes a communication connection with the target prediction model in real time; the user interface module updates the prediction results sent by the target prediction model in real time; and generates detailed report information containing predicted delivery time and mode of delivery from the prediction results.

[0185] It should be noted that in the above specific technical solution, the user interface module generates a detailed report containing the predicted time and mode of delivery in the mobile application. After the user interface module displays the predicted results, the pregnant woman can also make preparations in advance or actively adjust her life in the later stage of pregnancy, such as adjusting her diet, increasing exercise, etc.

[0186] Preferably, as an implementable method, after inputting the latest collected pregnancy data to be predicted, using the trained target prediction model to perform prediction processing operations on the delivery time and delivery mode, and outputting the delivery time and mode, it also includes:

[0187] Monitor the personal information of the pregnant women to be tested corresponding to the latest pregnancy data to be predicted;

[0188] Obtaining the medical history information of the pregnant woman to be tested extracted from the electronic health record system corresponding to the personal information of the pregnant woman to be tested;

[0189] Determine whether the pregnant woman to be tested currently has a history of chronic diseases. If so, mark the pregnant woman to be tested currently as a target pregnant woman for key attention.

[0190] Embodiment 2

[0191] Based on the same concept of the above method embodiment, the embodiment of the present invention also provides a processing system for predicting the time and mode of delivery based on data analysis based on machine learning, which is used to implement the above method of the present invention. Since the principles and methods of solving the problems in the system embodiment are similar, it has at least all the beneficial effects brought by the technical solutions of the above embodiments, which will not be repeated here one by one.

[0192] like Figure 7 As shown, the second embodiment of the present invention provides a processing system for predicting delivery time and delivery mode based on data analysis based on machine learning, including a data acquisition module 10, a data preprocessing module 20, a feature extraction module 30, a model training module 40 and a prediction module 50:

[0193] The data collection module 10 is used to collect pregnancy data;

[0194] A data preprocessing module 20, used to perform preprocessing operations on the collected pregnancy data to obtain a unified multi-source data set;

[0195] A feature extraction module 30 is used to extract features with a high correlation with the time and mode of delivery from the pre-processed unified multi-source data set and record them as target features;

[0196] A model training module 40 is used to train a machine learning model based on the selected target features to obtain a trained target prediction model;

[0197] The prediction module 50 is used to input the latest collected pregnancy data to be predicted, use the trained target prediction model to perform prediction processing operations on the delivery time and delivery mode, and output the delivery time and mode.

[0198] The specific architecture principle of the processing system for predicting delivery time and delivery mode based on machine learning data analysis is as follows:

[0199] 1. Data acquisition module; wearable device interface: use smart bracelets to monitor the heart rate, blood pressure and body movement data of pregnant women. Database connection of the hospital's electronic health record system: obtain the medical history of pregnant women, such as diabetes, hypertension, etc. from the hospital's electronic health record system. Ultrasound imaging data interface: obtain fetal development indicators, including fetal weight, head circumference, etc. through ultrasound equipment.

[0200] 2. Data preprocessing module; De-noising unit: Filter heart rate and blood pressure data to remove environmental noise interference. Data integration unit: Align data from different sources (wearable devices, hospital electronic health record system database, ultrasound imaging) by timestamp and integrate them into a unified data set.

[0201] 3. Feature extraction module; Statistical feature calculation unit: calculate the mean, standard deviation, maximum and minimum value of each physiological indicator. Correlation analysis unit: use the Pearson correlation coefficient to screen out features with a correlation with delivery time and mode higher than 0.7. Feature optimization unit: apply the recursive feature elimination (RFE) method to finally select the N most predictive features.

[0202] 4. Model training module; Machine learning algorithm selection unit: selects random forest algorithm as the main prediction model. Data set division unit: divides the data set into training set and test set in a ratio of 80:20. Model validation and adjustment unit: optimizes the number of trees and maximum depth parameters of random forest through 5-fold cross validation. Feature weight calculation unit performs calculation processing. Model input unit provides the calculated feature weight as input parameter to the random forest model. Model optimization unit: adjusts the internal parameters of random forest to minimize the prediction error.

[0203] 5. Prediction module; Data input unit: input the latest collected pregnancy data, including real-time monitored physiological indicators and the latest ultrasound data. Prediction calculation unit: use the trained random forest model to calculate the input data and generate the prediction result Y = W·X + b, where Y represents the predicted delivery time (in days) and delivery method (natural delivery or caesarean section). Result output unit: transmit the prediction result to the user interface module for display.

[0204] In summary, the present invention provides a processing method and system for predicting the time and mode of delivery based on data analysis based on machine learning, which realizes the full monitoring and utilization of pregnancy data of different pregnant women and production prediction through the integration of multi-source data, extraction of personalized features and accurate predictive analysis.

[0205] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. A person skilled in the art may modify the technical solutions described in the above embodiments, or replace part or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting delivery time and mode of delivery by implementing data analysis based on machine learning, characterized in that: The steps are as follows: Data collection during pregnancy: monitor the heart rate, blood pressure and body movement data of pregnant women through wearable device interfaces; extract the medical history information of pregnant women from the electronic health record system; obtain fetal development indicators through ultrasound equipment; Perform preprocessing operations on the collected pregnancy data to obtain a unified multi-source data set; Extract features with high correlation with delivery time and mode from the preprocessed unified multi-source data set and record them as target features; Use the machine learning model trained based on the selected target features to obtain a trained target prediction model; Input the latest collected pregnancy data to be predicted, use the trained target prediction model to predict the delivery time and mode of delivery, and output the delivery time and mode.

2. The method for predicting delivery time and delivery mode by data analysis based on machine learning according to claim 1, characterized in that: When the pregnancy data collection is executed, it also includes obtaining the physiological indicators of the pregnant woman through the wearable device interface; the physiological indicators of the pregnant woman include the heart rate, blood pressure and body movement data of the pregnant woman; the body movement data is the physical activity data recorded by the sensor device, including the number of steps walked per day and the length of sleep; The method also includes calling the database of the hospital's electronic health record system to obtain medical history information of multiple pregnant women; the medical history information of the pregnant women includes chronic disease history information; The method also includes calling a data interface of an ultrasonic device to obtain fetal development indicators; the fetal development indicators include fetal weight, head circumference, and abdominal circumference.

3. The method for predicting delivery time and delivery mode based on data analysis based on machine learning according to claim 1, characterized in that: The preprocessing operation performed on the collected pregnancy data to obtain a unified multi-source data set specifically includes: The denoising processing unit performs filtering processing on the collected heart rate, blood pressure data and body movement data to obtain filtered sampling data; The data integration unit aligns and integrates the filtered sampled data from wearable device monitoring, the database of the hospital's electronic health record system, and the ultrasound imaging data according to timestamps, and then outputs a unified multi-source data set.

4. The method for predicting delivery time and delivery mode based on data analysis based on machine learning according to claim 1, characterized in that: The features with high correlation with delivery time and mode are extracted from the preprocessed unified multi-source data set and recorded as target features, including: Statistical feature calculation unit: input the preprocessed unified multi-source data set, and then calculate the basic statistical features of each physiological indicator, including the mean, standard deviation, maximum value and minimum value; calculate the basic statistical features of each fetal developmental indicator, including the mean, standard deviation, maximum value and minimum value; summarize the basic statistical features of the above physiological indicators and the basic statistical features of the developmental indicators to obtain a statistical feature data set; Correlation analysis unit: input the statistical feature data set, use the Pearson correlation coefficient to evaluate the correlation between each basic statistical feature and the delivery time and mode of delivery, and screen out key features with a correlation coefficient higher than 0.7 as a high-correlation feature set; Feature optimization unit: Then input the high correlation feature set, apply the recursive feature elimination method, further optimize the key features, and finally select the N most predictive features as target features.

5. The method for predicting delivery time and delivery mode by implementing data analysis based on machine learning according to claim 4, characterized in that: The trained target prediction model is obtained by training the machine learning model based on the selected target features, including: The machine learning algorithm selection unit selects the random forest algorithm as the prediction model and records it as the random forest model; Then the data set partitioning unit partitions the preprocessed data set into a training set and a test set in a ratio of 80:20; The model validation and adjustment unit optimizes the key parameters of the random forest model through a 5-fold cross validation method; the key parameters include the number of trees (number of trees) and the maximum depth; Feature weight calculation unit: Based on the formula group F= { f 1， f 2’ … ， f 10} , calculate the weight of each feature, where The model input unit provides the calculated feature weights as input parameters to the random forest model, ensuring that the random forest model receives optimized feature data; The model optimization unit further adjusts the internal parameters of the random forest model to finally obtain a trained target prediction model; the internal parameters include sample weight allocation.

6. The method for predicting delivery time and delivery mode by data analysis based on machine learning according to claim 1, characterized in that: S105: input the latest collected pregnancy data to be predicted, use the trained target prediction model to perform prediction processing operations on the delivery time and mode of delivery, and output the delivery time and mode, specifically including: The data input unit inputs the latest collected pregnancy data to be predicted, wherein the input latest collected pregnancy data to be predicted includes physiological indicators of pregnant women monitored in real time and fetal development indicators obtained by the latest ultrasonic equipment; The prediction calculation unit uses the trained random forest model to calculate the input data and generate the prediction result Y=W·X+b; Wherein Y represents the predicted delivery time and delivery mode; the delivery time is in days; the delivery mode includes natural delivery or caesarean section; The result output unit transmits the prediction result to the user interface module, and displays the predicted time and mode of delivery to the pregnant woman and medical staff through the user interface module.

7. The method for predicting delivery time and delivery mode based on data analysis based on machine learning according to claim 6, characterized in that: The user interface module includes any one of a mobile application APP or an electronic health record system.

8. The method for predicting delivery time and delivery mode based on data analysis based on machine learning according to claim 7, characterized in that: The user interface module establishes a communication connection with the target prediction model in real time; the user interface module updates the prediction results sent by the target prediction model in real time; and generates detailed report information containing predicted delivery time and delivery mode from the prediction results.

9. The method for predicting delivery time and delivery mode by data analysis based on machine learning according to claim 7, characterized in that: After inputting the latest collected pregnancy data to be predicted, using the trained target prediction model to predict the delivery time and mode, and outputting the delivery time and mode, it also includes: Monitor the personal information of the pregnant women to be tested corresponding to the latest pregnancy data to be predicted; Obtaining the medical history information of the pregnant woman to be tested extracted from the electronic health record system corresponding to the personal information of the pregnant woman to be tested; Determine whether the pregnant woman to be tested currently has a history of chronic diseases. If so, mark the pregnant woman to be tested currently as a target pregnant woman for key attention.

10. A processing system for predicting delivery time and mode of delivery based on data analysis implemented by machine learning, characterized in that: Including data acquisition module, data preprocessing module, feature extraction module, model training module and prediction module: Data collection module, used to collect pregnancy data; A data preprocessing module is used to perform preprocessing operations on the collected pregnancy data to obtain a unified multi-source data set; A feature extraction module is used to extract features with a high correlation with the time and mode of delivery from the preprocessed unified multi-source data set and record them as target features; A model training module is used to train a machine learning model based on selected target features to obtain a trained target prediction model; The prediction module is used to input the latest collected pregnancy data to be predicted, use the trained target prediction model to perform prediction processing operations on the delivery time and mode of delivery, and output the delivery time and mode.

Images