Method and device for providing service for pregnant and puerperal users

Abstract

One or more embodiments of the invention provide a method and device for providing services for pregnant and puerperal users, and the method comprises the steps: obtaining user basic information, medical information and key protection information of a target user; the user basic information, the medical information and the key protection information are input into a pre-trained monitoring model, pregnancy and delivery health information of the target user is output and obtained, and the pregnancy and delivery health information is used for indicating a pregnancy and delivery type to which the target user belongs; wherein the monitoring model is a machine learning model obtained by taking user basic information, medical information and health care information of a sample user as training sample features and taking a risk index and a risk level corresponding to the sample user as training labels and carrying out supervised training.

Description

Technical Field

[0001] One or more embodiments of the present specification relate to the field of data processing technology, and more particularly, to a method and apparatus for providing services to pregnant and postpartum users. Background Art

[0002] In related technologies, before providing relevant services to pregnant and childbirth users, it is necessary to first create a birth card, generate a pregnancy and childbirth file, and record the user's pregnancy and childbirth information. Based on these files, we can provide more targeted services to pregnant and childbirth users. However, this card-building model relies too much on established files, resulting in users who have not created a card being unable to obtain relevant services, and unable to provide medical support in a timely manner based on unrecorded health information. Summary of the invention

[0003] In view of this, one or more embodiments of this specification provide the following technical solutions:

[0004] According to a first aspect of one or more embodiments of this specification, a method for providing services to pregnant and childbirth users is proposed, the method comprising:

[0005] Filtering health data of pregnant and postpartum users from the health data collection of users;

[0006] Based on the screening rules corresponding to the preset maternity type, the characteristic data corresponding to the maternity type in the health data of the maternity user is analyzed to obtain an analysis result; wherein the analysis result includes the maternity type to which the maternity user belongs;

[0007] Based on the analysis result, services related to the pregnancy and childbirth type are pushed to the pregnant and childbirth user.

[0008] According to a second aspect of one or more embodiments of this specification, a device for providing services to pregnant and postpartum users is provided, comprising:

[0009] A data collection module, used to filter the health data of pregnant and childbearing users from the health data set of users;

[0010] A data analysis module, configured to analyze the characteristic data corresponding to the maternity type in the health data of the maternity user based on the screening rules corresponding to the preset maternity type, and obtain an analysis result; wherein the analysis result includes the maternity type to which the maternity user belongs;

[0011] A service providing module is used to push services related to the pregnancy and childbirth type to the pregnancy and childbirth user based on the analysis results.

[0012] As can be seen from the above embodiments, this specification analyzes the health data of pregnant users through pre-set screening rules corresponding to the pregnancy type to determine their pregnancy type and provide targeted services accordingly. Unlike the traditional card-building model, this solution does not rely on pregnancy files, but actively screens health data to provide users with more timely and intelligent services, thereby protecting life and health and improving the quality of medical services. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 It is a schematic diagram of the architecture of a service system that provides services to pregnant and childbearing users, provided by an exemplary embodiment.

[0014] Figure 2 It is a flowchart of a method for providing services to pregnant and childbearing users provided by an exemplary embodiment.

[0015] Figure 3 It is a training diagram of a monitoring model provided by an exemplary embodiment.

[0016] Figure 4 It is a structural schematic diagram of a device provided by an exemplary embodiment.

[0017] Figure 5 It is a block diagram of a device for providing services to pregnant and childbearing users, provided by an exemplary embodiment. DETAILED DESCRIPTION

[0018] The user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this manual are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0019] Before providing relevant services to pregnant and childbirth users, it is necessary to first create a birth card, generate a pregnancy and childbirth file, and record the user's pregnancy and childbirth information. Based on these files, we can provide more targeted services to pregnant and childbirth users. However, this card-building model relies too much on established files, resulting in users who have not created a card being unable to obtain relevant services, and unable to provide medical support in a timely manner based on unrecorded health information.

[0020] In view of this, this specification proposes a method for providing services to pregnant and childbearing users, analyzing the health data of pregnant and childbearing users based on a pre-trained monitoring model to analyze and determine the pregnancy type of the pregnant and childbearing users, and thus pushing related services to the pregnant and childbearing users based on the pregnancy type.

[0021] During implementation, the health data of pregnant users are filtered from the user's health data set; based on the screening rules corresponding to the preset pregnancy type, the health data of the pregnant users are analyzed to obtain analysis results; wherein the analysis results include the pregnancy type to which the pregnant users belong; based on the analysis results, services related to the pregnancy type are pushed to the pregnant users.

[0022] In the above technical solution, the health data of pregnant users are analyzed through pre-set screening rules corresponding to the pregnancy type to determine their pregnancy type and provide targeted services accordingly. Unlike the traditional card-building model, this solution does not rely on pregnancy files, but actively screens health data to provide users with more timely and intelligent services, thereby protecting life and health and improving the quality of medical services.

[0023] Figure 1 FIG. 1 is a schematic diagram of the architecture of a service system for providing services to pregnant and childbirth users, provided by an exemplary embodiment. Figure 1 As shown, the system may include a server 11, a network 12, and several electronic devices, such as a PC (Personal Computer) 13, a mobile phone 14, and the like.

[0024] The server 11 may be a physical server including an independent host, or the server 11 may be a virtual server carried by a host cluster. During operation, the server 11 may run a server-side program of a certain application to implement the relevant functions of the application. For example, when the server 11 runs a service program that provides services to pregnant and childbirth users, it may be implemented as a corresponding service platform that provides services to pregnant and childbirth users.

[0025] PC13 and mobile phone 14 are only some types of electronic devices that users can use. In fact, users can obviously also use electronic devices such as the following types: tablet devices, laptops, PDAs (Personal Digital Assistants), wearable devices (such as smart glasses, smart watches, etc.), etc., and one or more embodiments of this specification do not limit this. During operation, the electronic device can run a program on the client side of a certain application to implement the relevant functions of the application. For example, when the electronic device runs a service program that provides services to pregnant and childbearing users, it can implement a service client that provides services to pregnant and childbearing users. Among them, the application program of the service client that provides services to pregnant and childbearing users can be started and run on the electronic device. The program on the client side can be a native application installed on the electronic device, or the program on the client side can be a small program, a quick application or other similar forms. Of course, when using web page technologies such as HTML5 or similar, the relevant functions can be implemented through the page displayed by the browser. The browser here can be an independent browser application or a browser module embedded in certain applications.

[0026] As for the network 12 for interaction between electronic devices such as PC 13 and mobile phone 14 and server 11, it is possible to select a wired or wireless network to achieve communication based on the communication mode supported by the corresponding electronic device, and this specification does not limit this. For example, PC 13 can support both wired and wireless communication, so it can use a wired or wireless network to achieve communication as needed, while mobile phone 14 usually only supports wireless communication, so it can use a wireless network to achieve communication.

[0027] In order to enable those skilled in the art to better understand the technical solutions in the present application, the technical solutions in this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0028] See also Figure 2 , Figure 2 A method for providing services to pregnant and childbearing users is provided by an exemplary embodiment. The method may include the following steps.

[0029] S201. Filter health data of pregnant and postpartum users from a collection of health data of users.

[0030] You can obtain a multi-dimensional health data set of users, which may include basic population information, medical information, public health information and other related data, for example, as shown in the following table:

[0031]

[0032]

[0033]

[0034] Based on the medical and health big data technology, it is possible to effectively interconnect health data and collect users' health data sets. By screening these data, the health data of pregnant and childbirth users can be extracted.

[0035] S202. Based on the screening rules corresponding to the preset pregnancy type, the characteristic data corresponding to the pregnancy type in the health data of the pregnant user is analyzed to obtain an analysis result; wherein the analysis result includes the pregnancy type to which the pregnant user belongs.

[0036] Several pregnancy and childbirth types can be determined in advance, and screening rules corresponding to each pregnancy and childbirth type can be set. The screening rules can be used to indicate the characteristic data corresponding to each pregnancy and childbirth type, the value range that needs to be met, and the rules for screening the characteristic data, etc.

[0037] Based on the screening rules corresponding to each pregnancy type, the screening rules corresponding to each pregnancy type in the health data of the pregnant users can be analyzed, and the obtained analysis results can be used to indicate the pregnancy type to which each pregnant user belongs.

[0038] The classification method of pregnancy and childbirth types can be set according to actual needs, and can be based on factors such as pregnancy status, card establishment status, and various risks that may exist during pregnancy. In one embodiment, the pregnancy and childbirth types may include: Eugenics and Two Free Users; Unestablished Card Users; High-risk Pregnant Women.

[0039] Among them, the "eugenics and two-free users" can be used to indicate the non-pregnant people who are suitable for pregnancy; the "unestablished card users" can be used to indicate the people who are pregnant and have not established a birth card; the "high-risk pregnant women" can be used to indicate the people who are pregnant and meet the risk screening conditions. Among them, the risk screening conditions may include screening conditions for several pregnancy risk indicators; if the pregnant and childbearing user meets the screening conditions of any pregnancy risk indicator, it can be confirmed that the pregnant and childbearing user is a high-risk pregnant and childbearing woman.

[0040] It should be noted that pregnant and childbearing users can belong to one or more types of pregnant and childbearing. For example, a pregnant and childbearing user can be both a non-card user and a high-risk pregnant and childbearing mother.

[0041] S203: Based on the analysis result, push services related to the pregnancy and childbirth type to the pregnant and childbirth user.

[0042] After obtaining the analysis results, the services that can be provided to pregnant users can be determined based on their pregnancy types. For example, for pregnant users who have not opened a birth card, they can be provided with card opening suggestions, pregnancy precautions, nearby relevant medical and health care institutions, etc., and relevant staff can be prompted to follow up; for high-risk pregnant users, medical advice and assisted medical treatment can be pushed to them, and medical staff can be prompted.

[0043] Based on the above embodiment, the health data of pregnant users are analyzed through pre-set screening rules corresponding to the pregnancy type to determine their pregnancy type and provide targeted services accordingly. Unlike the traditional card-building model, this solution does not rely on pregnancy files, but actively screens health data to provide users with more timely and intelligent services, thereby protecting life and health and improving the quality of medical services.

[0044] In one embodiment, a monitoring model for analyzing health data can be constructed based on screening rules corresponding to each maternity type, and the health data model obtained by training is obtained by taking the health data of a sample user as a training sample and taking the maternity type corresponding to the sample user as a training label. The health data of the maternity user is input into the pre-trained monitoring model for in-depth analysis, and the monitoring model screens the input health data and generates a series of analysis results, including determining the maternity type to which the maternity user belongs.

[0045] The monitoring model can be constructed according to actual needs, for example, based on neural networks, decision trees, clustering algorithms, etc. In one embodiment, a neural network based on the backpropagation (BP) algorithm can be used to construct the monitoring model. A large number of sample users' health data can be collected in advance, and data analysis can be performed on these health data to determine the characteristic data corresponding to different maternity types, and these characteristic data are used as the characteristics of the training samples. The corresponding analysis results including the maternity type are used as training labels to help the model learn the relationship between health data and maternity type. Through training, the monitoring model can identify the user's health characteristics and accurately determine their maternity type, which not only improves the accuracy of the analysis, but also lays the foundation for the subsequent provision of personalized health management and services.

[0046] Based on the above embodiments, the health data of pregnant and postpartum users is analyzed by using a pre-trained monitoring model, which can improve the efficiency and accuracy of the analysis.

[0047] In one embodiment, a preset timed task can be executed. Based on a preset time period, the health data of the pregnant and postpartum user can be regularly input into a pre-trained monitoring model for analysis. In each analysis process, a comprehensive analysis can also be performed in combination with historical analysis results.

[0048] In one embodiment, based on a preset time period, the health data of pregnant and postpartum users are regularly screened from the health data set of users, and the screened new health data is input into the monitoring model for analysis.

[0049] In one embodiment, after periodically screening the health data of pregnant and postpartum users from the user's health data set, it is possible to first determine whether the health data of the pregnant and postpartum users has been updated; if so, the updated health data needs to be input into the monitoring model for analysis; if not, there is no need to use the monitoring model to re-analyze the health data.

[0050] Among them, the time period can be set according to actual needs, for example, one week, one month, etc., and different time periods can be set according to different pregnancy types. For example, it can be set to one month for non-pregnant users, two weeks for pregnant users, one week for high-risk pregnant women, and so on.

[0051] By regularly analyzing the health data of pregnant and postpartum users, we can more timely understand changes in their health status and provide faster and more accurate services. This periodic data monitoring can not only help identify potential health problems, but also track various indicators of users during pregnancy and childbirth, providing pregnant and postpartum users with a comprehensive health monitoring system.

[0052] In one embodiment, since the data sources of various health data in the user's health data set are diverse, with characteristics such as large data volume, complex structure, inconsistent standards, and wide coverage, it is difficult to be fully utilized. Therefore, in order to facilitate the monitoring model to analyze the health data, the health data can be processed to generate structured data corresponding to the health data to improve the data quality of the health data; and then the structured data corresponding to the health data of pregnant and postpartum users is input into the monitoring model for analysis, which can improve the analysis efficiency of the monitoring model and the accuracy of the analysis results.

[0053] For example, in one implementation, big data governance technology can be used to collect, standardize, enrich, and perform privacy operations on massive amounts of health data.

[0054] Among them, methods for processing health data based on big data governance technology may include data standard management, data cleaning, indexing services, etc.

[0055] Data Standards Management:

[0056] Data standards management is the formulation and implementation of data standards, which mainly includes data set management, data element management, value range management, data verification standard management and master data management.

[0057] (1) Dataset management:

[0058] When conducting data set management, the first step is to conduct a comprehensive investigation of the data resources of each system. This includes sorting out the source, production method and other information of each data resource to form a detailed data ledger. Next, a classification framework is constructed based on the subject domain, such as statistical, construction and survey data sets, to form the primary, secondary and tertiary categories of the data set, and formulate corresponding coding rules.

[0059] Data element is the basic unit of data set. Data element management first needs to clarify its basic attributes, including identifier, name, definition, constraint, cardinality, representation format and allowed value. According to these attributes, the data elements in the data set are cataloged one by one, and finally a data element dictionary directory is generated. In the process of naming data elements, unified naming rules should be followed to ensure the uniqueness, simplicity, standardization and universality of the naming, so as to eliminate semantic ambiguity.

[0060] Through systematic data resource management, the availability and consistency of data can be improved, laying a solid foundation for subsequent data analysis and application.

[0061] (2) Range Management

[0062] The value domain is a set of allowed values ​​for a data element, and each allowed value consists of a specific value and its meaning. The value domain can be divided into two categories: enumerable value domain and non-enumerable value domain. For non-enumerable value domain, you need to use an accurate description to define its range, such as "a real number ≥0 and ≤1". For enumerable value domain, you should use the value domain management tool to list all the values ​​it contains and their corresponding meanings one by one.

[0063] By clearly defining and managing value ranges, data consistency and accuracy can be ensured, providing a solid foundation for subsequent data processing and analysis.

[0064] (3) Data Verification

[0065] The goal of data verification management is to build a comprehensive data verification indicator system, formulate data quality quantitative rules from multiple dimensions such as accuracy, completeness, timeliness and consistency, and ultimately form a data quality knowledge base.

[0066] Accuracy: Verify the validity of data element values, including whether they conform to the value domain definition and business logic.

[0067] Integrity: Verify the meaning of data element values ​​and check whether there are incomplete values ​​such as null values ​​(including invisible characters), meaningless values ​​(such as "0" in numeric format), or ignorable default values.

[0068] Timeliness: Verify the timeliness of data transmission and check whether the data is generated or transmitted on time through the agreed timeliness calculation rules.

[0069] Consistency: Verify the consistency between data sets and data elements, mainly checking whether the data records between the main data set and the sub-data set comply with the cardinality constraints to determine whether there is any data loss during data generation and transmission.

[0070] Master data is data that describes core business entities in health care and is used to integrate factual records related to these entities. When conducting master data management, you first need to screen out core business entities shared by multiple business systems. For each type of business entity, determine the only reliable data source, use publish and subscribe technology to monitor the dynamics of master data, and establish a synchronization and mapping management mechanism for master data.

[0071] Through the above measures, data quality can be effectively improved and the reliability and consistency of data can be ensured throughout its entire life cycle.

[0072] Data cleaning:

[0073] Data cleaning aims to deal with various problems in real-time collected health data, such as data ambiguity, duplication, incompleteness, and violation of business rules. Through conditional extraction, filtering, and screening, problematic data is sorted, verified, and corrected, while abnormal data that cannot be cleaned is diverted to ensure the availability, integrity, and consistency of data in the database. Data cleaning mainly includes the following aspects:

[0074] 1) Missing value processing

[0075] Detect missing values: Identify fields or records with missing values ​​in your data.

[0076] Fill missing values: Based on the characteristics of the data, use methods such as mean, median, mode, etc., or use interpolation to estimate.

[0077] Removing missing values: In certain cases, it is possible to remove records or fields that contain missing values, provided that this does not negatively impact the representativeness of the data.

[0078] 2) Outlier processing

[0079] Outlier detection: Identify outliers using statistical methods or machine learning algorithms (e.g. Z-Score, box plots).

[0080] Handling outliers: Based on business needs, outliers can be smoothed, replaced, or marked if necessary for subsequent processing.

[0081] 3) Duplicate value processing

[0082] Detect duplicate values: Identify whether there are identical records in the data.

[0083] Deduplication: Remove duplicate records to ensure data uniqueness.

[0084] 4) Standardization of formats

[0085] Unify field formats: Unify data from different sources, especially string data, into a consistent format to avoid problems caused by format differences.

[0086] Date and time formats: Standardize the representation format of dates and times to ensure correct interpretation during data analysis.

[0087] 5) Data type conversion

[0088] Check data types: Make sure the data type of each field meets the expected data type, such as converting string numbers to numeric types.

[0089] Type conversion: Convert data types as needed to meet the requirements of subsequent analysis or calculation.

[0090] 6) Uniqueness and consistency

[0091] Ensure uniqueness: Make sure identifiers or primary keys in your data are unique to avoid problems when linking and merging data.

[0092] Data consistency: ensuring that values ​​of fields representing the same information are consistent, such as standardization of country names.

[0093] 7) Data Verification

[0094] Business rule verification: Verify the legitimacy of data according to business rules to ensure compliance with business requirements.

[0095] Logical relationship verification: Check the logical relationship between data to ensure there is no contradiction.

[0096] 8) Data Normalization

[0097] Normalize data to standard units: For different units of measurement that have the same meaning, convert them to a unified standard unit.

[0098] Standardize numerical range: Standardizing numerical range helps to compare and analyze data at different scales.

[0099] Through the above steps, data cleaning can effectively improve data quality and provide a solid foundation for subsequent data analysis and decision-making.

[0100] Indexing Service:

[0101] The indexing service creates a unique identifier for the user through a specific algorithm, thereby achieving effective integration of the user's health data. The health data (such as outpatient records, prescriptions, doctor's orders, electronic medical records, examination and test reports, etc.) generated by each user when receiving health services at any medical institution in the region will be uniformly associated with their unique identifier. This association method forms a complete health data collection centered on the individual.

[0102] This complete collection of health data not only facilitates the tracking and management of health information, but also provides a reliable data foundation for subsequent screening and analysis. By integrating health data from different health institutions, medical service providers can have a more comprehensive understanding of the health status of residents, thereby achieving more accurate medical services and health management.

[0103] In one embodiment, health data may include a large amount of unstructured text data, such as clinical medical data, such as chief complaint, current medical history, past medical history, diagnostic information, medical records, etc., which contain a large amount of key personal health data. This type of health data can be semantically parsed using natural language processing technology to extract key text fragments representing semantics from the health data; based on the extracted key text fragments, structured data corresponding to the health data is constructed, which is equivalent to translating unstructured text data into structured data that can be understood by a computer, so as to provide data support for the analysis of the monitoring model, and the structured data corresponding to the health data of the pregnant user is input into the pre-trained monitoring model for analysis to determine the type of pregnancy and childbirth of the pregnant user.

[0104] Taking clinical medical data as an example, it can specifically include word-level language processing, syntactic-level language processing, and paragraph-level language processing:

[0105] Word-level analysis mainly includes Chinese word segmentation, named entity recognition, part-of-speech tagging, synonym segmentation, word vector, etc. For example, when a patient visits a doctor, the main complaint is described as: low fever and headache.

[0106] The syntactic-level analysis mainly includes dependency grammar analysis, word position analysis, semantic normalization, text error correction, etc. For example, when the patient comes to the hospital for treatment, the current medical history is described as: regular prenatal checkups during pregnancy, hypothyroidism during pregnancy, and drug treatment to date.

[0107] The analysis at the chapter level mainly includes label extraction, similarity analysis, topic model analysis, document classification and clustering. For example, when the patient was diagnosed, the physical examination description was: body temperature 37.8℃, blood pressure 127 / 69mmHg, pulse 105 times / min, respiration 14 times / min, oxygen saturation 100%, mental state, pregnant abdomen, no tenderness, previous surgical scars can be seen in the abdomen, fetal heart rate 166 times / min, internal examination did not check. Obstetric color Doppler ultrasound examination findings: fetal position: LOA, biparietal diameter: 47mm, femoral long diameter: 34mm, placenta: mainly anterior wall, thickness: 23mmGr: grade 0, sheep horizontal segment: 36mm, fetal heart rate: 163 times / min, regular, fetal movement: accessible, diagnosis opinion: intrauterine pregnancy single live fetus.

[0108] In one embodiment, a complete data processing flow for text data can be established. First, the original data is extracted from the text data, and the target data is formed through data cleaning, data type conversion, data normalization and other data processing operations as described above; then, based on the two main links of natural language processing and information extraction, the semantics in the target data is analyzed; the analysis results also need to be evaluated, applied, and fed back to achieve iterative optimization of the algorithm.

[0109] For example, a complete processing process for text data can be formed through data preparation, semantic analysis, evaluation feedback, and iterative optimization.

[0110] 1) Data preparation

[0111] Data extraction: Extract data-governed health data from the database to form original target data, laying the foundation for subsequent analysis.

[0112] Data preprocessing: Health data comes from various sources, and there are problems such as incompleteness and redundancy, which makes direct analysis difficult. Therefore, preprocessing is necessary to ensure the accuracy, completeness and consistency of the data. The quality of preprocessed data directly affects the calculation results, and high-quality data is more conducive to obtaining high-quality analysis results.

[0113] 2) Semantic analysis

[0114] Natural language processing: In order to effectively extract useful information from text data, it is necessary to apply natural language processing (NLP) technology to process unstructured text data.

[0115] Establish a semantic vocabulary: Based on manual annotation and machine learning, create a semantic vocabulary related to health data, mark and model medical terms, mathematical symbols, English abbreviations, etc., so that they can parse the meaning represented by semantics.

[0116] Information extraction: The entire text data is disassembled through the semantic word library, key information is extracted and relationship extraction is performed. For example, if the patient's diagnosis information is "pregnancy status" and the main complaint is "no desire to have children", after splitting and extracting through the semantic word library, it can be identified as:

[0117] Chief complaint: No desire to have children

[0118] Definition: Pregnant women and those without childbearing intentions. This will provide structured information for subsequent data applications.

[0119] 3) Evaluation feedback

[0120] Accuracy and precision evaluation: Compare the parsed text data with the sample data, calculate its accuracy and precision, analyze the differences and identify potential problems.

[0121] Business feedback and iteration: Collect feedback in business applications, continuously improve and iterate analysis algorithms to improve the accuracy and precision of semantic parsing. Through continuous improvement, ensure that the system can adapt to changing needs and improve the overall data processing effect.

[0122] In one implementation, feature data sets corresponding to each pregnancy and delivery type may be determined by performing feature association analysis on various health data of sample users.

[0123] There are many methods for performing feature association analysis, for example, a mining method based on a feature data set of relative risk.

[0124] Relative Risk (RR), also known as the hazard ratio, is the ratio of the hazard of the exposure group (the measurement indicator is the cumulative incidence) to the hazard of the control group. For example, taking a certain disease as an example, this indicator reflects the strength of the association between exposure (characteristic data) and the disease. The calculation formula is as follows: RR = cumulative incidence of exposure group / cumulative incidence of control group. For example, if you want to calculate whether dietary factors will cause complications during pregnancy, then the relationship table is as follows. The exposure group refers to those with irregular diet, the number of people with the disease is A, and the number of people without the disease is B; the control group refers to those with regular diet, the number of people with the disease is C, and the number of people without the disease is D, as shown in the following table:

[0125]

[0126]

[0127] The comparison of exposure factors and disease associations is shown in the following table:

[0128] RR value range Relevance 09-1.1 Not relevant 0.7-0.8 or 1.2-1.4 Weak association 0.4-0.6 or 1.5-2.9 Medium Association 0.1-0.3 or 3.0-9.9 Strong association <0.1 or >10 Strong correlation

[0129] Given a characteristic data of pregnancy complications D = {t1, t2, ..., t n}, each record is represented by a binary group t i =(x i ,y i ) to indicate that the second entry y i represents the disease, and the first entry x i represents the set of other attributes x i ={i1,i2,...,i m}.

[0130] For the sample data set, the relative risk is calculated as follows:

[0131]

[0132] The purpose of this algorithm is to conduct an in-depth analysis of factors (feature data) related to the disease, including both risk factors and protective factors. In view of the impact of these two aspects, the algorithm sets two threshold values: one is the risk threshold and the other is the protection threshold. The risk threshold is set to be greater than 1, and the protection threshold is between 0 and 1. Among them, the risk feature data set is a feature data set whose relative risk exceeds the critical value, and the safety feature data set is a feature data set whose relative risk is less than the safety critical value. These two types of feature data sets are collectively referred to as joint feature data sets. The core of this algorithm is to mine all associated feature data sets that meet the threshold requirements, and then construct the association rules of "{risk feature data set}=>disease" or "{protection feature data set}=>no disease".

[0133] In essence, both risk feature data sets and protection feature data sets conduct deeper analysis and processing of feature data, with different purposes and standards.

[0134] First, it is necessary to prepare the input data set D and set a risk threshold. This threshold is used to determine which attribute is risky. Generally, rr_risk will exceed 1. Secondly, we need to count the number of diseases, healthy numbers, and total numbers in the data set. These statistics will be used for later calculations. Then, the first round of scanning of the sample is performed to find feature data sets with risk coefficients greater than the risk threshold. These feature data are the initial results of risk identification, that is, candidate feature data. When a set of risk feature data sets is completed, they are cross-combined to obtain a potential risk feature data set. Here, k represents the length of the feature data. Scan the data again, but this time it is used to calculate the associated risk of each potential k item. Only when the relative risk is higher than the critical value can it be identified as an actual risk feature data set. This process will continue until no new candidate feature data can be generated. At this point, the mining work has ended.

[0135] The method of mining the characteristic data set of safety factors is similar to that of mining the risk characteristic data set. The biggest difference is that a protection threshold value greater than 0 and less than 1 must be set. The relative risk of a protection characteristic data set cannot be lower than this critical value.

[0136] In the above embodiment, a new correlation-based metric, namely relative risk, is proposed to replace the traditional Apriori algorithm. Relative risk refers to the difference between the proportion of cases in the exposure group and the control group, rather than the number of occurrences of a single factor. This method can effectively prevent the elimination of a small part in an unbalanced situation, and thus more accurately mine the implicit association rules.

[0137] In one implementation, a corresponding pruning method may also be used to improve the performance and convergence of feature association analysis when faced with large-scale data and massive rules.

[0138] However, there are limitations in mining the associated feature data sets of relative risk. When the number of feature data sets k increases, if there is no appropriate threshold limit, the number of feature data sets will increase sharply. The explosive growth of large-scale data will not only waste computing resources, but also generate a large number of association rules, which will impose a heavy burden on subsequent analysis and even cause data failure in serious cases. Therefore, it is necessary to prune it. When pruning, there are two main constraints: one is the superposition of effects, and the other is the number of exposure factor samples.

[0139] (1) Superposition of effects

[0140] When exploring the exposure factor feature data set, the k-1 feature data set can be extended to the k feature data set, and it is expected that the k feature data set has a better risk than the k-1 feature data set. This shows that when the exposure factors act together, their effects are positive, rather than interfering with each other. Therefore, while screening the candidate feature data set, it is also compared with the relative risk of the sub-feature data set.

[0141] (2) Number of exposure factor samples

[0142] Because relative risk is calculated using sampling ratios instead of limiting the sample size, there are situations where the relative risk is very high but the actual sample size is not statistically significant. For example, an exposure factor feature data set consists of {A, B, C}, but only 3 transactions are related to this feature data. Even if all three transactions are disease data, the RR value of the exposure factor feature data set obtained will be higher than the risk threshold. In this case, the conclusion is problematic and has no statistical significance, so it needs to be excluded. To address this issue, a pruning algorithm based on the number of samples is studied. This method sets a threshold for the minimum number of feature data sets. Only when the number of feature data sets exceeds this critical value, the feature data is considered to be statistically significant. In this way, feature data sets with too few samples or unrepresentative samples can be eliminated, making them more accurate and reliable.

[0143] In one embodiment, for pregnant women whose pregnancy type is not a card-opening user, corresponding screening rules can be determined based on various health data of the pregnant women, such as diagnostic information, medical history information, test reports, examination reports, medication information, maternal and child health information, basic population information, etc. Among them, since diagnostic information, medical history information, examination reports, etc. contain a large amount of unstructured text data, it is necessary to combine semantic analysis to extract effective information and generate structured text.

[0144] Since health data such as maternal medical information and fertility intentions will change dynamically over time, each analysis of health data needs to be combined with historical analysis results for a comprehensive analysis to obtain the analysis results.

[0145] The characteristic data corresponding to the unregistered user may include health data indicating the following information: whether the pregnant user is pregnant; whether the pregnant user has a desire to have a child; whether the pregnant user has abortion behavior. Accordingly, the obtained analysis results may include several scenarios as shown in the following table:

[0146] Scenario Historical health data Current health data 1 Not pregnant Pregnancy information (pregnant) 2 Not pregnant Pregnancy information + abortion or no fertility intention information 3 Not pregnant Information about abortion or lack of fertility 4 Pregnant Pregnancy information 5 Pregnant Pregnancy information + abortion or no fertility intention information 6 Pregnant Information about abortion or lack of fertility

[0147] After determining that the pregnant user is a non-card user, card creation suggestions, pregnancy precautions, nearby relevant medical care institutions, etc. can be pushed to the pregnant user, and relevant staff can be prompted to follow up. In one embodiment, different services can also be pushed based on different scenarios indicated by the analysis results.

[0148] In one embodiment, for pregnant users whose pregnancy type is high-risk, the health data of the pregnant users throughout their life cycle can be integrated, which may include: basic demographic information of the pregnant users (such as height, weight, age, dynamic BMI, etc.), historical health information (such as adverse pregnancy history, surgical history, drug abuse history, heart failure history, RH blood type, intellectual disability, etc.), pregnancy complication information (cardiovascular disease, respiratory system disease, digestive system disease, urinary system disease, endocrine system disease, blood system disease, nervous system disease, immune system disease, infectious disease, etc.), pregnancy complication information (threatened abortion, threatened premature birth, multiple pregnancy, abnormal amniotic fluid, low placenta, pregnancy complications, etc.), etc. Through the analysis of the integrated health data, several pregnancy risk indicators related to pregnant and childbearing users can be determined, and a dynamic comprehensive evaluation can be conducted based on the indicator weights (the weights can be dynamically adjusted according to the importance of the indicators) and the severity of the risks. According to the severity of the risks, several risk levels can be set. For example, they can be divided into 5 levels, with 5 colors as corresponding labels: "green (low risk), yellow (general risk), orange (higher risk), red (high risk), purple (infectious disease)".

[0149] As shown in the following table:

[0150]

[0151]

[0152]

[0153]

[0154]

[0155]

[0156]

[0157] As shown in the above table, the screening rules corresponding to high-risk pregnant women may include screening rules for various pregnancy risk indicators, and the corresponding characteristic data corresponding to high-risk pregnant women include characteristic data corresponding to various pregnancy risk indicators.

[0158] In one implementation, the analysis results obtained after analyzing the health data of pregnant users may include at least one pregnancy risk indicator and a risk level corresponding to the pregnancy risk indicator.

[0159] In one implementation, the analysis result obtained by analyzing the health data of a pregnant user, if it is determined that the pregnant user is a high-risk pregnant woman, the analysis result may also include the user risk level of the pregnant user.

[0160] Among them, the user risk level can be determined based on the risk level of the pregnancy risk indicator corresponding to the pregnant user. For example, the highest risk level among the risk levels of the pregnancy risk indicator corresponding to the pregnant user can be used as the user risk level of the pregnant user.

[0161] Figure 3 A method for constructing a monitoring model is shown, such as Figure 3 As shown, in the data collection stage, the health data of sample users can be obtained, including basic population information, medical information, health care information, etc.; in the data processing stage, the acquired health data is processed to form structured data, and the data processing stage may include: data standard management, data cleaning, indexing services, semantic analysis, etc.; the structured data is divided into a training data set and a test data set, and based on the screening rules corresponding to each pregnancy type, such as the screening rules for uncarded users, the screening rules for high-risk pregnant women, etc., a monitoring model for pregnant users is trained. The training process of the model may include the following steps:

[0162] (1) Model establishment

[0163] The BP algorithm for building a neural network consists of 1 input layer, several hidden layers and 1 output layer.

[0164] (2) Training Dataset

[0165] The input of the monitoring model can be the health data of sample users in the training data set. The health data is calculated and analyzed based on the screening rules. The analysis results output by the monitoring model can include the type of pregnancy and delivery.

[0166] (3) Data normalization

[0167] Data Normalization In order to make the network converge easily during training, the input data was normalized.

[0168] (4) Model algorithm

[0169] The output information of hidden layer 1 is obtained through the action of activation function f(u), and the output information of hidden layer 2 is obtained through the action of activation function f(u). The activation function f(u) adopts Sigmoid type, that is, f(u) = 1 / [1+exp(-u)]. The output information of the hidden layer is transmitted to the output layer to obtain the final output result.

[0170] (5) Algorithm iteration

[0171] If there is an error between the network output and the expected output, the error will be back-propagated by adjusting the network weights and thresholds: ①Δω(t) is the correction of the weights and thresholds during t trainings; ②η is called the learning rate, 0<η<1; ③E is the sum of squared errors and other parameters to optimize the algorithm's output to better fit the actual situation.

[0172] After training, the monitoring model can be used to actively screen the health data of pregnant and childbearing users. The evaluation criteria for the risk level of pregnancy risk indicators can be shown in the following table:

[0173]

[0174]

[0175]

[0176]

[0177]

[0178]

[0179]

[0180]

[0181] By collecting and processing a large amount of health data, screening rules and monitoring models can be established to analyze the multi-dimensional health data of pregnant and childbearing users, and obtain the analysis results such as the pregnancy type of the pregnant and childbearing users, as well as pregnancy risk indicators and corresponding risk levels, and display them in an intuitive form. For example, different colors can be used to distinguish different risk levels.

[0182] By dividing and labeling people with similar health characteristics and behavioral patterns, managers can quickly identify user groups with characteristic pregnancy and delivery types, thereby providing more accurate services.

[0183] Individualized interventions can be developed for different maternity groups:

[0184] (1) Active intervention

[0185] One-stop service integration: Promote cooperation between civil affairs departments and health departments, and complete premarital and pre-pregnancy examinations simultaneously during marriage registration to ensure that every newlywed couple can receive necessary medical guidance.

[0186] Free e-voucher promotion: Distribute free e-vouchers for Eugenics through mobile Internet platforms to encourage more people to participate. In addition, detailed graphic descriptions are provided to explain the content of each test and its importance.

[0187] Full-process companion service: Optimize the existing physical examination process, disassemble and reorganize 19 pre-examination and pregnancy examination items according to the three nodes of before, during and after the examination. Each link has corresponding digital guidance to help users complete it smoothly. At the same time, the information system is used to automatically push the physical examination results, and online communication and feedback are supported to enhance the user experience.

[0188] (2) Monitoring and early warning

[0189] Screening of users without a card: Using the data analysis method described above, we can screen out women who are pregnant but have not yet obtained a maternity card and prenatal care handbook from the health data of pregnant and childbearing users, and promptly notify relevant medical staff to pay attention to and follow up on the corresponding situation.

[0190] Establish a five-color early warning system: For pregnant women who are identified as pregnant, five colors (green, yellow, orange, red, and purple) are used to represent different levels of risk. This allows doctors to check the patient's specific situation (including but not limited to high-risk factors, previous medical records, etc.) at any time and adjust their risk level based on the latest progress. Find problems as early as possible and take corresponding measures to reduce the probability of adverse consequences.

[0191] Figure 4 is a schematic structural diagram of a device provided by an exemplary embodiment. Figure 4 At the hardware level, the device includes a processor 402, an internal bus 404, a network interface 406, a memory 408, and a non-volatile memory 410, and may also include hardware required for other functions. One or more embodiments of this specification may be implemented based on software, such as the processor 402 reading the corresponding computer program from the non-volatile memory 410 into the memory 408 and then running it. Of course, in addition to the software implementation, one or more embodiments of this specification do not exclude other implementations, such as logic devices or a combination of software and hardware, etc., that is, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0192] Please refer to Figure 5 , the device for providing services to pregnant and childbirth users can be applied to Figure 4 The device shown in the figure is used to implement the technical solution of this specification. The device for providing services to pregnant and childbirth users may include: a data collection module 501, a data analysis module 502 and a service provision module 503.

[0193] The data collection module 501 is used to filter the health data of pregnant and childbearing users from the health data set of users;

[0194] The data analysis module 502 is used to analyze the characteristic data corresponding to the maternity type in the health data of the maternity user based on the screening rules corresponding to the preset maternity type, and obtain the analysis result; wherein the analysis result includes the maternity type to which the maternity user belongs;

[0195] The service providing module 503 is used to push services related to the pregnancy and childbirth type to the pregnancy and childbirth user based on the analysis results.

[0196] In one embodiment, the data analysis module 502 is used to input the health data of the pregnant user into a pre-trained monitoring model for analysis, and obtain the analysis results output by the monitoring model, wherein the analysis results include the pregnancy type to which the pregnant user belongs; wherein the monitoring model is a model obtained by training using the feature data corresponding to the pregnancy type in the health data of the sample user as the features of the training sample, and using the analysis results corresponding to the sample user as training labels.

[0197] In one embodiment, the data analysis module 502 is used to execute a preset timed task, and based on a preset time period, regularly input the health data of the pregnant and postpartum user into a pre-trained monitoring model for analysis.

[0198] In one implementation, the data collection module 501 is further used to process the health data to generate structured data corresponding to the health data;

[0199] The data analysis module 502 is used to input the structured data corresponding to the health data of the pregnant and postpartum user into a pre-trained monitoring model for analysis.

[0200] In one embodiment, the health data includes text data;

[0201] The data analysis module 502 is used to perform semantic analysis on the health data to extract key text segments representing semantics from the health data; and to construct structured data corresponding to the health data based on the extracted key text segments.

[0202] In one embodiment, the pregnancy type includes:

[0203] Eugenics two-free users, the Eugenics two-free users are used to indicate the non-pregnant population of suitable pregnancy;

[0204] Uncarded users, which are used to indicate people who are pregnant but have not established a fertility card;

[0205] High-risk pregnant women, wherein the high-risk pregnant women are used to indicate people who are pregnant and meet risk screening conditions.

[0206] In one implementation, the pregnancy type is a user without a pregnancy card, and the characteristic data corresponding to the user without a pregnancy card is used to indicate the following information:

[0207] Whether the pregnant user is pregnant;

[0208] Whether the pregnant user has the desire to have children;

[0209] Whether the pregnant user has had an abortion.

[0210] In one embodiment, the pregnancy type is high-risk pregnancy, the characteristic data corresponding to the high-risk pregnancy includes characteristic data corresponding to several pregnancy risk indicators, and the analysis result also includes at least one pregnancy risk indicator and a risk level corresponding to the at least one risk indicator.

[0211] Based on the same concept as the above method, this specification also provides an electronic device, including: a processor; a memory for storing processor executable instructions; wherein the processor implements the steps of the method described in any of the above embodiments by running the executable instructions.

[0212] Based on the same concept as the above method, this specification also provides a computer-readable storage medium on which computer instructions are stored. When the instructions are executed by a processor, the steps of the method described in any of the above embodiments are implemented.

[0213] Based on the same concept as the above method, this specification also provides a computer program product, including a computer program / instruction, which implements the steps of the method described in any of the above embodiments when executed by a processor.

Claims

1. A method for providing services to pregnant and childbearing users, characterized in that: The method comprises: Filtering health data of pregnant and postpartum users from the health data collection of users; Based on the screening rules corresponding to the preset maternity type, the characteristic data corresponding to the maternity type in the health data of the maternity user is analyzed to obtain an analysis result; wherein the analysis result includes the maternity type to which the maternity user belongs; Based on the analysis result, services related to the pregnancy and childbirth type are pushed to the pregnant and childbirth user.

2. The method according to claim 1, characterized in that The health data of the pregnant user is analyzed based on the screening rules corresponding to the preset pregnancy type to obtain the analysis results, including: The health data of the pregnant user is input into a pre-trained monitoring model for analysis, and the analysis results output by the monitoring model are obtained, wherein the analysis results include the pregnancy type to which the pregnant user belongs; wherein the monitoring model is a model obtained by training using the feature data corresponding to the pregnancy type in the health data of the sample user as the features of the training sample, and using the analysis results corresponding to the sample user as training labels.

3. The method according to claim 1, characterized in that: The step of inputting the health data of the pregnant and postpartum user into a pre-trained monitoring model for analysis includes: Execute the preset timed tasks, and based on the preset time period, regularly input the health data of the pregnant and postpartum user into the pre-trained monitoring model for analysis.

4. The method according to claim 1, characterized in that Before the health data is input into the pre-trained monitoring model for analysis and calculation, it also includes: Processing the health data to generate structured data corresponding to the health data; Inputting the health data of the pregnant and postpartum user into a pre-trained monitoring model for analysis, including: The structured data corresponding to the health data of the pregnant and postpartum user is input into a pre-trained monitoring model for analysis.

5. The method according to claim 4, characterized in that The health data includes text data; The processing of the health data to generate structured data corresponding to the health data includes: Performing semantic analysis on the health data to extract key text segments representing semantics from the health data; Build structured data corresponding to health data based on the extracted key text fragments.

6. The method according to any one of claims 1 to 5, characterized in that: The types of pregnancy include: Eugenics two-free users, the Eugenics two-free users are used to indicate the non-pregnant population of suitable pregnancy; Uncarded users, which are used to indicate people who are pregnant but have not established a fertility card; High-risk pregnant women, wherein the high-risk pregnant women are used to indicate people who are pregnant and meet risk screening conditions.

7. The method according to claim 6, characterized in that The pregnancy type is a user without a pregnancy card. The characteristic data corresponding to the user without a pregnancy card is used to indicate the following information: Whether the pregnant user is pregnant; Whether the pregnant user has the desire to have children; Whether the pregnant user has had an abortion.

8. The method according to claim 6, characterized in that The pregnancy type is high-risk pregnant women, and the characteristic data corresponding to the high-risk pregnant women include characteristic data corresponding to several pregnancy risk indicators. The analysis result also includes at least one pregnancy risk indicator and a risk level corresponding to the at least one risk indicator.

9. A device for providing services to pregnant and childbirth users, characterized in that: include: A data collection module, used to filter the health data of pregnant and childbearing users from the health data set of users; A data analysis module, configured to analyze the characteristic data corresponding to the maternity type in the health data of the maternity user based on the screening rules corresponding to the preset maternity type, and obtain an analysis result; wherein the analysis result includes the maternity type to which the maternity user belongs; A service providing module is used to push services related to the pregnancy and childbirth type to the pregnancy and childbirth user based on the analysis results.

10. An electronic device, characterized in that: include: processor; A memory for storing processor-executable instructions; wherein the processor implements the steps of the method according to any one of claims 1 to 8 by executing the executable instructions.

11. A computer-readable storage medium, characterized in that: Computer instructions are stored thereon, and when the instructions are executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

12. A computer program product, characterized in that The method comprises a computer program / instruction, which, when executed by a processor, implements the steps of the method according to any one of claims 1 to 8.

Images