Information recommendation method and system for medical and health data mining

Abstract

The present invention relates to the field of medical information mining technology, and in particular to an information recommendation method and system for medical health data mining. The method comprises the following steps: acquiring multi-source medical data; extracting medical keywords from the multi-source medical data to obtain medical information keyword data; constructing a medical knowledge graph for standard multi-source medical data using the medical information keyword data to obtain a multi-source medical knowledge graph; acquiring user physiological health information data; performing abnormal physiological hazard assessment on the user physiological health information data to generate user physiological abnormal hazard data; importing the user physiological abnormal hazard data into the multi-source medical knowledge graph to perform first medical health data mining, thereby generating first medical health recommendation data. The present invention improves the accuracy and personalization of medical health data mining information recommendations through multi-source data acquisition, personalized recommendation, dynamic adjustment and real-time update.

Description

Technical Field

[0001] The present invention relates to the technical field of medical information mining, and in particular to an information recommendation method and system applied to medical health data mining. Background Art

[0002] In recent years, with the explosive growth of medical and health data, the application of information recommendation methods in medical and health data mining has become increasingly important. Traditional medical data mainly comes from medical records, laboratory test results, and imaging data. The data volume is relatively small, and the data mining methods are relatively simple. With the popularization of electronic health records (EHR), the types and scale of medical data have increased significantly, and data mining technology has ushered in new challenges and opportunities. Platforms based on big data technology such as Hadoop and Spark are increasingly widely used in medical data processing. These platforms can efficiently process and analyze large-scale medical data, improving the efficiency and accuracy of data mining and information recommendation. At the same time, with the advancement of natural language processing (NLP) technology, the mining ability of medical text data has been significantly improved, and important information in medical records and literature can be more accurately extracted and used. However, at present, traditional medical and health data mining often ignores individual differences, lacks personalization of recommendation results, and lacks the ability to dynamically adjust according to user feedback, resulting in poor recommendation results, resulting in low accuracy and personalization of health data mining information recommendation. Summary of the invention

[0003] Based on this, it is necessary to provide an information recommendation method and system for medical and health data mining to solve at least one of the above technical problems.

[0004] To achieve the above object, an information recommendation method for medical health data mining is provided, the method comprising the following steps:

[0005] Step S1: Acquire multi-source medical data; extract medical keywords from the multi-source medical data to obtain medical information keyword data; construct a medical knowledge graph for the standard multi-source medical data using the medical information keyword data to obtain a multi-source medical knowledge graph;

[0006] Step S2: Acquire user physiological health information data; perform abnormal physiological hazard assessment on the user physiological health information data to generate user physiological abnormal hazard data; import the user physiological abnormal hazard data into a multi-source medical knowledge graph to perform first medical health data mining, thereby generating first medical health recommendation data;

[0007] Step S3: collecting recommendation feedback information of users based on the first medical health recommendation data to obtain user feedback information data; dynamically adjusting the score recommendation of the first medical health recommendation data according to the user feedback information data, thereby generating second medical health recommendation data;

[0008] Step S4: access the second medical and health recommendation data in real time to generate a real-time medical recommendation data stream; construct personalized recommendation decisions on the real-time medical recommendation data stream to generate a medical and health recommendation decision stream; dynamically publish the medical and health recommendation decision stream to a terminal to generate a real-time interface for medical and health information recommendation.

[0009] The present invention can efficiently integrate and associate multiple medical data by constructing a multi-source medical knowledge graph, accurately mine the relevant medical health information of the user's physiological abnormalities, and provide accurate first medical health recommendation data. The user's physiological health information and feedback data are used to dynamically adjust the medical health recommendation content, thereby generating second medical health recommendation data that is more in line with the individual needs of the user, and realizing personalized medical services. Through real-time data stream access and personalized recommendation decision construction, the system can quickly respond to user needs, generate medical health recommendation decision streams in real time, and dynamically publish them to the terminal to improve user experience and satisfaction. Based on user feedback data, the recommended content is continuously optimized, and the effectiveness and user acceptance of the recommended content are guaranteed through score recommendation and dynamic adjustment, thereby continuously improving the recommendation effect of the system. Containerized processing and real-time data stream technology are used to ensure the efficiency and stability of the system when processing a large amount of medical data, and personalized medical health information is displayed through an intuitive dynamic recommendation interface, which is convenient for users to obtain and understand relevant health advice. Multi-source medical data and user personalized data are comprehensively utilized to systematically and intelligently manage and recommend medical health information, improve the overall medical health management level, and help users to carry out scientific health management and prevention. Therefore, the present invention improves the accuracy and personalization of medical and health data mining information recommendation through multi-source data acquisition, personalized recommendation, dynamic adjustment and real-time update.

[0010] Preferably, step S1 comprises the following steps:

[0011] Step S11: Acquire multi-source medical data;

[0012] Step S12: performing data preprocessing on the multi-source medical data to generate standard multi-source medical data, wherein the data preprocessing includes data cleaning, data missing value filling and data standardization;

[0013] Step S13: extracting medical keywords from standard multi-source medical data using natural language technology to obtain medical information keyword data;

[0014] Step S14: Use medical information keyword data to construct a medical knowledge graph for standard multi-source medical data to obtain a multi-source medical knowledge graph.

[0015] By acquiring and preprocessing multi-source medical data, the present invention can unify the management of medical data from different sources and in different formats, eliminate the phenomenon of data islands, and improve the overall quality and consistency of the data. Through data cleaning and missing value filling, the integrity and accuracy of the data are ensured, the analysis deviation caused by missing or incorrect data is reduced, and the utilization value of the data is improved. The data standardization step converts data from different sources into a unified standard format, which is convenient for subsequent analysis and application, reduces the complexity of data processing, and improves the operability and analysis efficiency of the data. By using natural language technology to extract medical keywords from standardized data, key information in the data can be quickly and accurately extracted, providing an important basis for the construction of a medical knowledge graph. By constructing a multi-source medical knowledge graph, knowledge in the medical field can be systematically organized and represented to form a structured knowledge network, which is convenient for deeper knowledge association and mining. The construction of a multi-source medical knowledge graph helps to realize the visualization and association analysis of medical information, reveal the complex relationship between data, and support clinical decision-making and medical research.

[0016] Preferably, step S14 includes the following steps:

[0017] Step S141: semantically analyzing the medical information keyword data to obtain medical information semantic relationship data;

[0018] Step S142: performing semantic relationship recognition on the medical information keyword data through the medical information semantic relationship data to generate a medical information semantic relationship matrix;

[0019] Step S143: extracting entity names from standard multi-source medical data based on the medical information semantic relationship matrix to obtain medical information entity naming data;

[0020] Step S144: perform homonym disambiguation on the medical information entity naming data to generate a standardized medical entity set; construct a medical ontology based on the medical information keyword data and the standardized medical entity set to generate a medical ontology; perform entity association on the standardized medical entity set through the medical ontology to generate a multi-source medical knowledge graph.

[0021] The present invention can identify and extract semantic relationships in data, mine deep information hidden in data, and improve the ability to understand data by semantically parsing medical information keyword data. Using semantic relationship data to generate a semantic relationship matrix helps to accurately identify various semantic relationships in medical information, establish associations between data, and form a structured semantic network. Based on the semantic relationship matrix, entity naming extraction can automatically identify and name entities in medical information, form standardized medical entity data, and lay the foundation for the construction of knowledge graphs. Medical information entity naming data is disambiguated with homonyms to eliminate ambiguity and duplication in the data, generate standardized medical entity sets, and improve the accuracy and consistency of data. Medical ontology is constructed through medical information keyword data and standardized medical entity sets to form a systematic and structured medical knowledge system, which is convenient for knowledge organization and retrieval. Medical ontology is used to perform entity association on standardized medical entity sets to generate multi-source medical knowledge graphs, realize the association and fusion between data, and enhance the utilization value of data.

[0022] Preferably, step S2 comprises the following steps:

[0023] Step S21: Acquire user physiological health information data;

[0024] Step S22: identifying abnormal physiological health parameters of the user's physiological health information data to obtain abnormal physiological parameters of the user;

[0025] Step S23: performing a hazard assessment on the abnormal physiological parameters of the user to generate the abnormal physiological hazard data of the user;

[0026] Step S24: Import the user's physiological abnormality hazard data into the multi-source medical knowledge graph to perform first medical health data mining, thereby generating first medical health recommendation data.

[0027] The present invention can comprehensively monitor the health status of users and timely discover physiological abnormal parameters by acquiring physiological health information data of users. Abnormal physiological health parameters are identified for physiological health information data of users, and the health abnormalities of users can be detected quickly and accurately, ensuring the timeliness and accuracy of health monitoring. The harmfulness assessment of abnormal physiological parameters of users is performed to generate abnormal physiological hazard data of users, which is helpful to evaluate the potential impact of abnormal parameters on the health of users and provide a scientific basis for formulating personalized health plans. The abnormal physiological hazard data of users is imported into a multi-source medical knowledge graph for first medical health data mining, and rich medical knowledge resources can be used for in-depth analysis to improve the accuracy and comprehensiveness of data mining. The abnormal physiological hazard data of users is imported into a multi-source medical knowledge graph for first medical health data mining, and rich medical knowledge resources can be used for in-depth analysis to improve the accuracy and comprehensiveness of data mining. The abnormal physiological hazard data of users is integrated with the multi-source medical knowledge graph to promote data sharing and application, and improve the utilization efficiency and value of medical information. Through personalized health recommendations, the health management experience of users is improved, helping users to better understand and manage their health status, and increasing the participation and satisfaction of users.

[0028] Preferably, step S23 includes the following steps:

[0029] Step S231: performing time series change analysis on the abnormal physiological parameters of the user to generate abnormal physiological change data of the user; performing fast Fourier transform based on the abnormal physiological change data of the user to generate a spectrum diagram of abnormal physiological change of the user; performing frequency domain analysis on the spectrum diagram of abnormal physiological change of the user to generate frequency domain data of abnormal physiological change of the user;

[0030] Step S232: performing health stability assessment on the frequency domain data of abnormal physiological changes of the user to generate physiological health stability data of the user;

[0031] Step S233: performing an abnormality over-limit analysis on the user's physiological abnormality parameters to generate user physiological abnormality over-limit data; performing an event-induced association analysis on the user's physiological abnormality over-limit data according to a multi-source medical knowledge graph to generate event-induced association data;

[0032] Step S234: using the user's physiological health stability data to assign weights to the event-induced association data based on the harmfulness of the associated events, and obtaining the user's physiological abnormality hazard data.

[0033] The present invention can identify the changing trend of abnormal parameters over time and reveal the potential health risk dynamics by analyzing the time series changes of abnormal physiological parameters of users. Based on the fast Fourier transform, the spectrum of abnormal physiological changes of users is generated, and the frequency domain analysis is performed to identify the periodicity and frequency characteristics of abnormal parameters, providing a scientific basis for in-depth understanding of the nature of abnormal parameters. The health stability assessment of the frequency domain data of abnormal physiological changes of users is performed to generate the physiological health stability data of users, which is helpful to evaluate the stability and fluctuation of the overall health status of users and guide personalized health management. The abnormal over-limit analysis of the abnormal physiological parameters of users is performed to identify and quantify the abnormal parameters beyond the normal range, provide detailed abnormal characteristic data, and provide a basis for subsequent health risk assessment. Based on the multi-source medical knowledge graph, the event-induced association analysis of the abnormal physiological over-limit data of users is performed to reveal the correlation between abnormal parameters and potential induced events, which is helpful to understand the cause and background of the abnormality. Using the physiological health stability data of users, the event-induced association data is weighted for the harmfulness of the associated events, and the physiological abnormal hazard data of users is obtained, which provides scientific health risk assessment results and supports personalized health management decisions. Through time series changes and frequency domain analysis, it is possible to dynamically monitor changes in the user's health status, provide real-time health risk warnings and management recommendations, and improve the timeliness and effectiveness of health management.

[0034] Preferably, step S24 includes the following steps:

[0035] Step S241: confirming the hazard impact factor of the user's physiological abnormality hazard data to obtain the physiological abnormality hazard impact factor;

[0036] Step S242: importing physiological abnormality hazard influencing factors into the multi-source medical knowledge graph to perform hazard keyword search to obtain hazard keyword associated data; performing association classification on the multi-source medical knowledge graph according to the hazard keyword associated data to generate hazard overlap associated data and hazard non-overlap associated data;

[0037] Step S243: Prioritize the mining of hazard overlap-related data and hazard non-overlap-related data to generate association mining priority data; perform first medical health data mining on the user's physiological abnormality hazard data through the association mining priority data to generate first medical health recommendation data.

[0038] The present invention can identify and quantify the impact of abnormal parameters on health by confirming the hazard influencing factors of the user's physiological abnormal hazard data, provide a scientific basis, and provide guidance for subsequent health management. The physiological abnormal hazard influencing factors are imported into the multi-source medical knowledge graph for hazard keyword retrieval, which can quickly find relevant medical knowledge and data, and improve the efficiency and accuracy of data retrieval. The multi-source medical knowledge graph is associated and classified according to the hazard keyword association data, and hazard overlap association data and hazard non-overlap association data are generated, which can identify the relationship and commonality between abnormal parameters, which helps to deeply understand health risks. The hazard overlap association data and the hazard non-overlap association data are mined and the association mining priority data is generated, which helps to determine the focus and order of data mining and improve the efficiency and effect of data mining. The first medical health data mining of the user's physiological abnormal hazard data is carried out through the association mining priority data, which can accurately mine the key data related to the user's health and generate scientific and effective medical health recommendation data. Based on the first medical health recommendation data obtained by mining, personalized health management suggestions and plans are provided to users, the pertinence and effectiveness of health management are improved, and the health experience of users is enhanced. Through multi-level hazard assessment and knowledge graph mining, it provides scientific and comprehensive health risk assessment results to support medical professionals in formulating precise health management plans.

[0039] Preferably, step S3 comprises the following steps:

[0040] Step S31: collecting recommendation feedback information of users based on the first medical health recommendation data to obtain user feedback information data; performing user feedback satisfaction detection on the user feedback information data to generate user feedback satisfaction detection data;

[0041] Step S32: recommending and scoring the first medical health recommendation data according to the user feedback satisfaction test data, and generating positive recommendation score data and negative recommendation score data;

[0042] Step S33: hiding the related recommendations of the first medical health recommendation data according to the negative recommendation score data, thereby generating medical health negative recommendation optimization data; dynamically adjusting the recommended items of the first medical health recommendation data based on the positive recommendation score data, thereby generating medical health positive recommendation optimization data;

[0043] Step S34: Integrate the medical health negative recommendation optimization data and the medical health positive recommendation optimization data to generate second medical health recommendation data.

[0044] The present invention collects user recommendation feedback information based on the first medical health recommendation data, and can timely understand the user's actual experience and feedback on health advice, providing a basis for subsequent optimization. Perform user feedback satisfaction detection on the user feedback information data, generate user feedback satisfaction detection data, quantify user satisfaction, and evaluate the effect of the recommendation scheme. Recommendation score is performed on the first medical health recommendation data through the user feedback satisfaction detection data, and positive recommendation score data and negative recommendation score data are generated, so that effective and recommended items that need to be improved can be identified. According to the negative recommendation score data, the first medical health recommendation data is hidden for related recommendations, and medical health negative recommendation optimization data is generated, which is helpful to eliminate the recommended items that the user is dissatisfied with or invalid, and improve the accuracy of the recommendation and user experience. Based on the positive recommendation score data, the first medical health recommendation data is dynamically adjusted for the recommended items, and medical health positive recommendation optimization data is generated, which can dynamically adjust the priority of the recommended items to ensure that the user gets the most relevant and effective health advice. The medical health negative recommendation optimization data and the medical health positive recommendation optimization data are integrated to generate the second medical health recommendation data, and a comprehensive optimized health management suggestion is provided to improve the overall quality of the recommendation. Through real-time analysis and optimization of user feedback, more accurate and personalized health management suggestions are provided to improve users' health management experience and satisfaction. Based on user feedback and dynamic adjustment mechanisms, the health recommendation system can be continuously improved to ensure the update and optimization of recommendation data and improve the scientificity and effectiveness of recommendations.

[0045] Preferably, step S31 includes the following steps:

[0046] Step S311: collecting recommendation feedback information of the user based on the first medical health recommendation data to obtain user feedback information data, wherein the user feedback information data includes user interaction data and user direct feedback data;

[0047] Step S312: performing user interaction intensity analysis on the user interaction data to generate user interaction intensity data; performing feedback sentiment analysis on the user direct feedback data to generate user direct feedback sentiment data;

[0048] Step S313: Perform recommendation scoring based on the user's direct feedback emotion data and user interaction intensity data to generate a health recommendation index;

[0049] Step S314: Use the health recommendation index to perform user feedback satisfaction detection on the first medical health recommendation data to generate user feedback satisfaction detection data.

[0050] The present invention collects user recommendation feedback information based on the first medical health recommendation data to obtain user feedback information data, which can comprehensively collect user interaction data and direct feedback data to ensure the diversity and integrity of feedback information. Performing user interaction intensity analysis on user interaction data to generate user interaction intensity data can quantify the user's interaction in the recommendation system and reflect the user's attention to and use of the recommended content. Performing feedback sentiment analysis on user direct feedback data to generate user direct feedback sentiment data can identify the user's emotional attitude towards the recommended content and provide more detailed feedback information. Performing recommendation scoring through user direct feedback sentiment data and user interaction intensity data to generate a health recommendation index can comprehensively evaluate the user's satisfaction with the recommended content and provide a more accurate evaluation index. Using the health recommendation index to perform user feedback satisfaction detection on the first medical health recommendation data to generate user feedback satisfaction detection data can comprehensively and accurately evaluate the user's satisfaction with the recommended content and provide a scientific basis for subsequent optimization. Through comprehensive user feedback analysis, the advantages and disadvantages of the recommended content can be identified, targeted optimization can be performed, and the quality of the recommended content and user satisfaction can be improved. Combining user interaction intensity and sentiment analysis, we can more accurately understand users' personalized needs and preferences and provide more targeted health management suggestions. By collecting and analyzing user feedback from multiple aspects, we can provide recommendations that better meet user needs and improve user experience and satisfaction.

[0051] Preferably, step S4 comprises the following steps:

[0052] Step S41: Containerizing the second medical health recommendation data to generate a health data containerization recommendation model;

[0053] Step S42: using the health data containerization recommendation model to access the second medical health recommendation data in real-time data stream, thereby generating a real-time medical recommendation data stream;

[0054] Step S43: constructing personalized recommendation decisions on the real-time medical recommendation data stream by implementing a decision engine to generate a medical health recommendation decision stream;

[0055] Step S44: Dynamically recommend terminal publishing of the medical and health recommendation decision flow, thereby generating a real-time interface for medical and health information recommendation.

[0056] The present invention improves the flexibility and portability of data processing by containerizing the second medical health recommendation data and generating a health data containerization recommendation model, thereby ensuring the efficient operation and scalability of the system. The health data containerization recommendation model is used to access the second medical health recommendation data in real-time data stream to generate a real-time medical recommendation data stream, which can update and process the health recommendation data in real time to ensure the timeliness and accuracy of the recommended content. By implementing a decision engine to construct a personalized recommendation decision for the real-time medical recommendation data stream and generate a medical health recommendation decision stream, personalized health management suggestions can be provided according to the user's real-time data and needs, and the relevance and user satisfaction of the recommended content can be improved. The medical health recommendation decision stream is dynamically recommended to the terminal release, and a real-time interface for medical health information recommendation is generated, which can display health management suggestions in real time, so that users can view and obtain the latest health information at any time. Through real-time data stream access and dynamic recommendation release, it is possible to respond to user needs and changes in real time, provide timely health management suggestions, and improve user experience and compliance. Containerization improves the flexibility and portability of the system, and can quickly deploy and update the health recommendation system to ensure the efficient operation of the system. Through the construction of personalized recommendation decisions, targeted health management suggestions can be provided according to the actual situation and needs of users, improving the accuracy and effectiveness of recommended content. Through dynamic recommendation publishing, users can obtain health management suggestions in real time, which improves user convenience and satisfaction and enhances user experience. Using real-time data streams and decision engines, data-driven personalized recommendation decisions can be made to ensure the scientificity and effectiveness of recommended content.

[0057] In this specification, an information recommendation system for medical and health data mining is provided, which is used to execute the above-mentioned information recommendation method for medical and health data mining. The information recommendation system for medical and health data mining includes:

[0058] The knowledge graph construction module is used to obtain multi-source medical data; extract medical keywords from the multi-source medical data to obtain medical information keyword data; use the medical information keyword data to construct a medical knowledge graph for standard multi-source medical data to obtain a multi-source medical knowledge graph;

[0059] A shallow medical mining module is used to obtain user physiological health information data; perform abnormal physiological hazard assessment on the user's physiological health information data to generate user physiological abnormal hazard data; import the user's physiological abnormal hazard data into a multi-source medical knowledge graph to perform first medical health data mining, thereby generating first medical health recommendation data;

[0060] A deep medical mining module is used to collect recommendation feedback information from users based on the first medical health recommendation data to obtain user feedback information data; dynamically adjust the score recommendation of the first medical health recommendation data according to the user feedback information data, thereby generating second medical health recommendation data;

[0061] The decision delivery module is used to access the real-time data stream of the second medical and health recommendation data to generate a real-time medical recommendation data stream; construct personalized recommendation decisions for the real-time medical recommendation data stream to generate a medical and health recommendation decision stream; and dynamically recommend the terminal to publish the medical and health recommendation decision stream to generate a real-time interface for medical and health information recommendation.

[0062] The beneficial effect of the present invention is that by acquiring multi-source medical data and user physiological health information data, the system can integrate multiple data sources to obtain more comprehensive and accurate medical information. By constructing a knowledge graph for medical information keyword data, the system can organize medical knowledge into a structured form, providing more efficient and convenient knowledge acquisition and recommendation. Abnormal assessment of user physiological health information data can help identify potential physiological abnormality risks. Through dynamic adjustment of user feedback information data and score recommendations, the system can continuously optimize the recommendation results according to the user's personalized needs and feedback, and provide recommendation data that is more in line with the user's health status and preferences. By constructing personalized recommendation decisions for real-time medical recommendation data streams and generating a real-time interface for medical health information recommendations, the system can provide real-time, personalized medical health information recommendation services to help users obtain and apply relevant medical knowledge and health advice in a timely manner. Therefore, the present invention improves the accuracy and personalization of medical health data mining information recommendations through multi-source data acquisition, personalized recommendations, dynamic adjustments and real-time updates. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 A schematic diagram of the steps of an information recommendation method applied to medical and health data mining;

[0064] Figure 2 for Figure 1 Detailed implementation steps of step S2 in the flowchart;

[0065] Figure 3 for Figure 1 Detailed implementation steps of step S3 in FIG.

[0066] Figure 4 for Figure 1 Detailed implementation steps of step S4 in FIG.

[0067] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0068] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are 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 technicians in this field without creative work are within the scope of protection of the present invention.

[0069] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0070] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0071] To achieve this, please refer to Figures 1 to 4 , an information recommendation method applied to medical health data mining, the method comprising the following steps:

[0072] Step S1: Acquire multi-source medical data; extract medical keywords from the multi-source medical data to obtain medical information keyword data; construct a medical knowledge graph for the standard multi-source medical data using the medical information keyword data to obtain a multi-source medical knowledge graph;

[0073] Step S2: Acquire user physiological health information data; perform abnormal physiological hazard assessment on the user physiological health information data to generate user physiological abnormal hazard data; import the user physiological abnormal hazard data into a multi-source medical knowledge graph to perform first medical health data mining, thereby generating first medical health recommendation data;

[0074] Step S3: collecting recommendation feedback information of users based on the first medical health recommendation data to obtain user feedback information data; dynamically adjusting the score recommendation of the first medical health recommendation data according to the user feedback information data, thereby generating second medical health recommendation data;

[0075] Step S4: access the second medical and health recommendation data in real time to generate a real-time medical recommendation data stream; construct personalized recommendation decisions on the real-time medical recommendation data stream to generate a medical and health recommendation decision stream; dynamically publish the medical and health recommendation decision stream to a terminal to generate a real-time interface for medical and health information recommendation.

[0076] The present invention can efficiently integrate and associate multiple medical data by constructing a multi-source medical knowledge graph, accurately mine the relevant medical health information of the user's physiological abnormalities, and provide accurate first medical health recommendation data. The user's physiological health information and feedback data are used to dynamically adjust the medical health recommendation content, thereby generating second medical health recommendation data that is more in line with the individual needs of the user, and realizing personalized medical services. Through real-time data stream access and personalized recommendation decision construction, the system can quickly respond to user needs, generate medical health recommendation decision streams in real time, and dynamically publish them to the terminal to improve user experience and satisfaction. Based on user feedback data, the recommended content is continuously optimized, and the effectiveness and user acceptance of the recommended content are guaranteed through score recommendation and dynamic adjustment, thereby continuously improving the recommendation effect of the system. Containerized processing and real-time data stream technology are used to ensure the efficiency and stability of the system when processing a large amount of medical data, and personalized medical health information is displayed through an intuitive dynamic recommendation interface, which is convenient for users to obtain and understand relevant health advice. Multi-source medical data and user personalized data are comprehensively utilized to systematically and intelligently manage and recommend medical health information, improve the overall medical health management level, and help users to carry out scientific health management and prevention. Therefore, the present invention improves the accuracy and personalization of medical and health data mining information recommendation through multi-source data acquisition, personalized recommendation, dynamic adjustment and real-time update.

[0077] In the embodiment of the present invention, reference Figure 1 The above is a schematic flow chart of the steps of an information recommendation method applied to medical health data mining of the present invention. In this example, the information recommendation method applied to medical health data mining includes the following steps:

[0078] Step S1: Acquire multi-source medical data; extract medical keywords from the multi-source medical data to obtain medical information keyword data; construct a medical knowledge graph for the standard multi-source medical data using the medical information keyword data to obtain a multi-source medical knowledge graph;

[0079] In an embodiment of the present invention, data is obtained from multiple medical data sources such as hospitals, health examination centers, laboratories, and patient health management systems, including but not limited to electronic medical records (EMR), imaging data, laboratory test results, patient health records, wearable device data, etc. Process the noise, redundant information, and erroneous records in the data to ensure the accuracy and completeness of the data. Mean filling, interpolation, or similarity-based filling methods are used to fill missing values ​​in the data. The data is normalized to uniformly convert data from different sources and formats into a standard format. Preprocessing steps such as word segmentation, part-of-speech tagging, stop word removal, and stem extraction are performed on the text data. Important medical keywords are extracted from the text data using methods such as the TF-IDF (term frequency-inverse document frequency) algorithm, the TextRank algorithm, or the BERT model based on deep learning. Remove irrelevant or repeated keywords, retain keywords with medical significance, and store the extracted medical keyword data in a database for subsequent processing. Use natural language processing technology to perform semantic analysis on medical information keywords and generate medical information semantic relationship data. Through the medical information semantic relationship data, the semantic relationship of medical information keywords is identified to generate a medical information semantic relationship matrix. Based on the semantic relationship matrix, entity naming is extracted from standard multi-source medical data to obtain medical information entity naming data. Medical information entity naming data is disambiguated with homonyms to generate a standardized medical entity set. Medical ontology is constructed based on medical information keyword data and standardized medical entity set to generate a medical ontology. Entity association is performed on the standardized medical entity set through the medical ontology to generate a multi-source medical knowledge graph. Define the types and relationships of nodes (such as diseases, symptoms, treatment methods, drugs, etc.) and edges (such as disease-symptoms, disease-treatment methods, etc.) of the medical knowledge graph. Add relevant attributes and relationships to each node and edge to enrich the information of the graph. Use graph database tools such as Neo4j and RDF to generate a multi-source medical knowledge graph, and store the generated knowledge graph in the graph database for subsequent query and use.

[0080] Step S2: Acquire user physiological health information data; perform abnormal physiological hazard assessment on the user physiological health information data to generate user physiological abnormal hazard data; import the user physiological abnormal hazard data into a multi-source medical knowledge graph to perform first medical health data mining, thereby generating first medical health recommendation data;

[0081] In an embodiment of the present invention, the physiological health information data of the user is obtained from multiple sources such as the user's wearable device, health examination report, medical record, real-time monitoring device, etc., including but not limited to heart rate, blood pressure, body temperature, blood oxygen saturation, respiratory rate, exercise volume, sleep quality and other physiological health parameters. The baseline value of the normal range is set according to medical standards and the user's historical health data, and the physiological parameters beyond the normal range are identified by statistical analysis and machine learning algorithms (such as abnormal detection algorithm, cluster analysis, etc.), so as to obtain the abnormal physiological parameters of the user. The abnormal physiological parameters of the user are analyzed in time series, the changing trend of the parameters is observed, and the abnormal physiological change data of the user is generated. The abnormal physiological change data of the user is fast Fourier transformed to generate the abnormal physiological change spectrum of the user. The abnormal physiological change spectrum of the user is analyzed in frequency domain to generate the abnormal physiological change frequency domain data of the user. The abnormal physiological change frequency domain data of the user is evaluated for health stability to generate the physiological health stability data of the user. The abnormal physiological parameters of the user are analyzed in excess of the limit to identify the abnormal parameters exceeding the safety threshold and generate the abnormal physiological excess limit data of the user. Using the multi-source medical knowledge graph, the event-induced association analysis of the user's physiological abnormality over-limit data is performed to generate event-induced association data. Based on the user's physiological health stability data, the event-induced association data is weighted by the harmfulness of the associated events to generate the user's physiological abnormality harm data. The generated user physiological abnormality harm data is imported into the multi-source medical knowledge graph, and the knowledge graph is updated and maintained. The harm influencing factors of the user's physiological abnormality harm data are confirmed, and the physiological abnormality harm influencing factors are extracted. The physiological abnormality harm influencing factors are imported into the multi-source medical knowledge graph, and harm keywords are searched to obtain harm keyword associated data. According to the harm keyword associated data, the multi-source medical knowledge graph is associated and classified to generate harm overlap association data and harm non-overlap association data. The harm overlap association data and the harm non-overlap association data are mined and prioritized to generate associated mining priority data. Through the associated mining priority data, the user's physiological abnormality harm data is mined for the first medical health data, thereby generating the first medical health recommendation data.

[0082] Step S3: collecting recommendation feedback information of users based on the first medical health recommendation data to obtain user feedback information data; dynamically adjusting the score recommendation of the first medical health recommendation data according to the user feedback information data, thereby generating second medical health recommendation data;

[0083] In an embodiment of the present invention, user interaction data is collected through user behaviors (such as clicks, browsing, interaction time, etc.) on a health management application or platform. User direct feedback data is collected through feedback forms, comments, questionnaires, etc. filled out by users. User interaction data is preprocessed, including data cleaning, deduplication, etc. The intensity of the user in the interaction process (such as the number of clicks, dwell time, interaction frequency, etc.) is calculated using statistical analysis methods to generate user interaction intensity data. Text analysis is performed on the user direct feedback data, and the user's emotional tendency (positive, negative, neutral) is extracted using a sentiment analysis algorithm (such as a sentiment dictionary, a machine learning model, etc.) to generate user direct feedback sentiment data. Combining the user direct feedback sentiment data and the user interaction intensity data, a scoring model (such as weighted average, linear regression, etc.) is used to calculate the health recommendation index of each recommendation. The user feedback satisfaction test data is analyzed to evaluate the user's satisfaction with each recommendation, and the user feedback satisfaction test data is generated. According to the satisfaction test results, the recommendation score is divided into positive recommendation score data and negative recommendation score data. The negative recommendation score data is processed, and the recommendations with lower scores are hidden or deleted to generate medical health negative recommendation optimization data. Process the positive recommendation score data, adjust the recommendation items and display order, give priority to the recommendations with higher scores, and generate medical and health positive recommendation optimization data. Merge and optimize the medical and health negative recommendation optimization data and the medical and health positive recommendation optimization data to ensure the accuracy of the recommendations and user experience, and finally generate the second medical and health recommendation data.

[0084] Step S4: access the second medical and health recommendation data in real time to generate a real-time medical recommendation data stream; construct personalized recommendation decisions on the real-time medical recommendation data stream to generate a medical and health recommendation decision stream; dynamically publish the medical and health recommendation decision stream to a terminal to generate a real-time interface for medical and health information recommendation.

[0085] In an embodiment of the present invention, by using containerization technologies such as Docker or Kubernetes, the second medical health recommendation data is packaged into an independent container image to generate a containerized recommendation model for health data. The container is deployed and managed through a container orchestration tool to ensure that the recommendation model can be quickly started and run in different environments. The second medical health recommendation data is accessed and processed in real time using a real-time data stream processing framework such as Apache Kafka, Apache Flink or Spark Streaming to generate a real-time medical recommendation data stream. The accessed real-time data is cleaned, converted and filtered to ensure the validity and accuracy of the data. Using Drools or other rule engines, the real-time medical recommendation data stream is processed and decided by defining a series of business rules and logic. Based on the user's historical data, behavior data and preference data, combined with the real-time medical recommendation data stream, a personalized medical health recommendation decision stream is generated. The medical health recommendation decision stream is published to the user's terminal device, such as a mobile phone, tablet, computer, etc., to ensure that the user can receive personalized health recommendation information in real time. Design and develop a dynamic recommendation real-time interface to display medical and health recommendation information, including charts, text, and interactive elements to improve user experience. More specifically, use Docker to create a container image and package the second medical and health recommendation data and recommendation algorithm into the container. Deploy the container image to the Kubernetes cluster to realize the automated deployment and management of the container. Configure the Kafka cluster to publish and subscribe the second medical and health recommendation data as a Kafka topic. Create a real-time data stream processing task in Flink or SparkStreaming to receive data in the Kafka topic for processing and analysis. Define decision rules in Drools to make decisions based on user historical data and real-time data streams. Implement a personalized recommendation algorithm to generate a personalized recommendation decision stream by combining real-time data streams and decision rules. Develop a front-end interface to display medical and health recommendation information. The interface includes interactive elements such as charts, text, and buttons. Implement a data push mechanism to push the medical and health recommendation decision stream to the user terminal to achieve real-time updates.

[0086] Preferably, step S1 comprises the following steps:

[0087] Step S11: Acquire multi-source medical data;

[0088] Step S12: performing data preprocessing on the multi-source medical data to generate standard multi-source medical data, wherein the data preprocessing includes data cleaning, data missing value filling and data standardization;

[0089] Step S13: extracting medical keywords from standard multi-source medical data using natural language technology to obtain medical information keyword data;

[0090] Step S14: Use medical information keyword data to construct a medical knowledge graph for standard multi-source medical data to obtain a multi-source medical knowledge graph.

[0091] In an embodiment of the present invention, the source of multi-source medical data is determined, such as a hospital information system, a health record platform, a mobile health application, etc. The data acquisition method is determined, involving API interfaces, data file import, etc. The data obtained from different sources are integrated and stored to ensure the integrity and consistency of the data. A data warehouse or database management system needs to be established to store these data. Errors, duplications, incomplete or invalid data in the data are cleared to ensure data quality. Outlier detection and processing are performed to ensure the reliability and accuracy of the data. The missing values ​​in the data are analyzed and appropriate filling strategies are adopted, such as mean filling, median filling or model-based filling methods. Ensure that no deviation or distortion is introduced during the filling process. The data is converted into a unified format and unit for subsequent processing and analysis. It involves normalization or standardization of the numerical range to ensure the comparability and consistency of the data during the processing process. The text information in the medical data is processed using natural language processing (NLP) technology. Including technologies such as word segmentation, part-of-speech tagging, and entity recognition, key information and keywords are extracted from the text. Keywords and terms related to medical care are extracted based on professional knowledge and corpus in the medical field. You can use techniques such as bag-of-words model and TF-IDF to identify and weight the importance of keywords. Based on the medical information keyword data, build a medical knowledge graph to describe the relationship between medical entities (such as diseases, treatment methods, drugs, etc.). Use graph databases or other technologies and tools suitable for knowledge graph construction, such as Neo4j. Determine the relationship model between medical entities, such as the relationship between diseases and symptoms, the relationship between treatment methods and diseases, etc. Build a complete knowledge graph model to support subsequent data query and analysis. Update the medical knowledge graph regularly to keep it synchronized with the latest medical knowledge. Optimize the graph structure and query efficiency to improve the performance and response speed of data retrieval.

[0092] Preferably, step S14 includes the following steps:

[0093] Step S141: semantically analyzing the medical information keyword data to obtain medical information semantic relationship data;

[0094] Step S142: performing semantic relationship recognition on the medical information keyword data through the medical information semantic relationship data to generate a medical information semantic relationship matrix;

[0095] Step S143: extracting entity names from standard multi-source medical data based on the medical information semantic relationship matrix to obtain medical information entity naming data;

[0096] Step S144: perform homonym disambiguation on the medical information entity naming data to generate a standardized medical entity set; construct a medical ontology based on the medical information keyword data and the standardized medical entity set to generate a medical ontology; perform entity association on the standardized medical entity set through the medical ontology to generate a multi-source medical knowledge graph.

[0097] In an embodiment of the present invention, the medical information keyword data is deeply parsed by using natural language processing (NLP) technology, such as dependency syntax analysis, semantic role labeling, etc. These tools can identify grammatical structures and semantic relationships in sentences. Semantic relationships between medical information, such as treatment, diagnosis, symptoms, etc., are extracted from the parsing results. Based on the semantic parsing results in step S141, a medical information semantic relationship matrix is ​​established. Each item of the matrix represents the strength or type of the semantic relationship between two medical information entities. According to the semantic relationship matrix, medical information entities with semantic relationships are extracted from standard multi-source medical data. The names and boundaries of the entities are determined for further processing and association. Semantic similarity calculation or context-based methods are used to solve the problem of homonymy to ensure that the generated standardized medical entity set is accurate. The knowledge base or semantic network can be combined to define and classify entities more accurately. A medical ontology is constructed based on the medical information keyword data and the standardized medical entity set. The medical ontology includes the hierarchical structure, attributes and relationships of medical entities, which are used to more deeply understand and describe the knowledge in the medical field. The standardized medical entity set is associated with the medical ontology to establish the relationship between the entities. These relationships can be hierarchical, associative, or semantically dependent and are used to build a multi-source medical knowledge graph.

[0098] As an example of the present invention, refer to Figure 2 As shown, in this example, step S2 includes:

[0099] Step S21: Acquire user physiological health information data;

[0100] Step S22: identifying abnormal physiological health parameters of the user's physiological health information data to obtain abnormal physiological parameters of the user;

[0101] Step S23: performing a hazard assessment on the abnormal physiological parameters of the user to generate the abnormal physiological hazard data of the user;

[0102] Step S24: Import the user's physiological abnormality hazard data into the multi-source medical knowledge graph to perform first medical health data mining, thereby generating first medical health recommendation data.

[0103] In an embodiment of the present invention, the user's physiological health information data, such as heart rate, blood pressure, blood sugar, body temperature, etc., is obtained from various sensor devices or health monitoring devices. A data acquisition system is designed and deployed to ensure real-time data acquisition from sensor devices. Considering the real-time and accuracy of the data, appropriate communication protocols and data transmission technologies, such as Bluetooth, Wi-Fi, or dedicated sensor networks, are used. Using machine learning or statistical analysis methods, an algorithm is developed to identify abnormal parameters in the user's physiological health data. Supervised learning, unsupervised learning, or semi-supervised learning techniques can be used to perform abnormality detection based on historical data and thresholds of health indicators. The identified abnormal parameters are classified, such as hypertension, arrhythmia, diabetes risk, etc. Based on medical knowledge and health indicators, the potential harm of each abnormal parameter to the user's health is evaluated. Considering factors such as the duration, severity, and possible complications of the abnormal parameters, detailed hazard assessment data is generated. Hazards are graded, such as mild, moderate, and severe, for subsequent medical management. The generated user physiological abnormal hazard data is integrated into a multi-source medical knowledge graph. Ensure the consistency and accuracy of the data, and associate the data with the existing medical knowledge graph through standardization and matching mechanisms. Use the data and relationships in the knowledge graph to mine and analyze medical data. Through data mining algorithms and models, potential health risks, disease associations, or health recommendations can be discovered.

[0104] Preferably, step S23 includes the following steps:

[0105] Step S231: performing time series change analysis on the abnormal physiological parameters of the user to generate abnormal physiological change data of the user; performing fast Fourier transform based on the abnormal physiological change data of the user to generate a spectrum diagram of abnormal physiological change of the user; performing frequency domain analysis on the spectrum diagram of abnormal physiological change of the user to generate frequency domain data of abnormal physiological change of the user;

[0106] Step S232: performing health stability assessment on the frequency domain data of abnormal physiological changes of the user to generate physiological health stability data of the user;

[0107] Step S233: performing an abnormality over-limit analysis on the user's physiological abnormality parameters to generate user physiological abnormality over-limit data; performing an event-induced association analysis on the user's physiological abnormality over-limit data according to a multi-source medical knowledge graph to generate event-induced association data;

[0108] Step S234: using the user's physiological health stability data to assign weights to the event-induced association data based on the harmfulness of the associated events, and obtaining the user's physiological abnormality hazard data.

[0109] In an embodiment of the present invention, the time series data of the user's physiological health parameters are obtained, including the time series data of parameters such as blood pressure, heart rate, and blood sugar. The time series data is analyzed to identify abnormal change patterns of the user's physiological health parameters, such as sudden increase, sudden decrease, or continuous fluctuation. The time series data is applied to the fast Fourier transform algorithm to convert the time domain data into frequency domain data. A spectrum diagram of abnormal physiological changes of the user is generated to show the abnormal physiological changes at different frequencies, such as high-frequency and low-frequency components. The frequency domain data is analyzed to evaluate the physiological health stability of the user, considering the distribution and change trend of the frequency components. Statistical methods or machine learning algorithms can be used to quantify indicators of the user's physiological health stability, such as the mean, variance, and power spectral density of frequency domain features. The abnormal change data is compared with the preset health range or standard to identify abnormal data points that exceed the normal range. The abnormal over-limit data is associated with the relevant medical knowledge in the multi-source medical knowledge graph for association analysis. The association between abnormal data diseases or health events is explored, such as heart attack, diabetes exacerbation, etc. Based on the user's physiological health stability data, the potential harm of event-induced associated data to the user's health is evaluated. The severity of the event, possible consequences, and the user's current physiological state are considered to assign a harm weight.

[0110] Preferably, step S24 includes the following steps:

[0111] Step S241: confirming the hazard impact factor of the user's physiological abnormality hazard data to obtain the physiological abnormality hazard impact factor;

[0112] Step S242: importing physiological abnormality hazard influencing factors into the multi-source medical knowledge graph to perform hazard keyword search to obtain hazard keyword associated data; performing association classification on the multi-source medical knowledge graph according to the hazard keyword associated data to generate hazard overlap associated data and hazard non-overlap associated data;

[0113] Step S243: Prioritize the mining of hazard overlap-related data and hazard non-overlap-related data to generate association mining priority data; perform first medical health data mining on the user's physiological abnormality hazard data through the association mining priority data to generate first medical health recommendation data.

[0114] In an embodiment of the present invention, by analyzing the physiological abnormality hazard data of the user, the key factors and parameters affecting the health of the user are determined. The specific influencing factors of physiological abnormality on the health of the user can be confirmed by statistical analysis, machine learning algorithms or expert experience. The confirmed physiological abnormality hazard influencing factors are used as keywords and imported into the multi-source medical knowledge graph for retrieval. Medical knowledge and information associated with these keywords are obtained, including diseases, symptoms, treatment methods, etc. According to the hazard keyword association data, the information in the multi-source medical knowledge graph is classified. Information directly related to physiological abnormality hazards (hazard overlap association data) and indirectly related or unrelated information (hazard non-overlap association data) are distinguished. Priority is assigned to hazard overlap association data and hazard non-overlap association data according to the importance of the information and the degree of association with physiological abnormality hazards. Methods such as weight allocation and association scoring can be used to generate priority data for association mining for subsequent data mining and analysis processes. Based on the priority data, the physiological abnormality hazard data of the user is deeply mined and analyzed. Combined with the relevant information in the multi-source medical knowledge graph, the first medical health recommendation data for the user's health status is generated, including preventive measures, treatment suggestions or health management plans.

[0115] As an example of the present invention, refer to Figure 3 As shown, in this example, step S3 includes:

[0116] Step S31: collecting recommendation feedback information of users based on the first medical health recommendation data to obtain user feedback information data; performing user feedback satisfaction detection on the user feedback information data to generate user feedback satisfaction detection data;

[0117] Step S32: recommending and scoring the first medical health recommendation data according to the user feedback satisfaction test data, and generating positive recommendation score data and negative recommendation score data;

[0118] Step S33: hiding the related recommendations of the first medical health recommendation data according to the negative recommendation score data, thereby generating medical health negative recommendation optimization data; dynamically adjusting the recommended items of the first medical health recommendation data based on the positive recommendation score data, thereby generating medical health positive recommendation optimization data;

[0119] Step S34: Integrate the medical health negative recommendation optimization data and the medical health positive recommendation optimization data to generate second medical health recommendation data.

[0120] In an embodiment of the present invention, feedback information is collected from the user according to the first medical health recommendation data. The feedback information may include user satisfaction, suggestions, evaluations, etc., which are obtained through the interaction between the user and the recommendation system. The user feedback information data is analyzed to evaluate the user's satisfaction with the recommendation system. Natural language processing technology can be used to perform sentiment analysis on text feedback, or satisfaction evaluation can be performed through user behavior data. According to the user's negative feedback score or sentiment analysis results, the recommended content that the user is dissatisfied with or recommends hiding is identified. For these negative scoring data, the relevant recommended items are hidden or the recommendation frequency is reduced to improve the user experience. According to the user's positive feedback score or sentiment analysis results, the recommended items and frequency of the first medical health recommendation data are adjusted and optimized. Improve the user's satisfaction and acceptance of the recommended content, and ensure the rationality and personalization of the recommended content. The data after negative recommendation optimization and positive recommendation optimization are integrated. The result of the integration is the second medical health recommendation data, which has been adjusted and optimized by user feedback to provide recommended content that better meets user needs.

[0121] Preferably, step S31 includes the following steps:

[0122] Step S311: collecting recommendation feedback information of the user based on the first medical health recommendation data to obtain user feedback information data, wherein the user feedback information data includes user interaction data and user direct feedback data;

[0123] Step S312: performing user interaction intensity analysis on the user interaction data to generate user interaction intensity data; performing feedback sentiment analysis on the user direct feedback data to generate user direct feedback sentiment data;

[0124] Step S313: Perform recommendation scoring based on the user's direct feedback emotion data and user interaction intensity data to generate a health recommendation index;

[0125] Step S314: Use the health recommendation index to perform user feedback satisfaction detection on the first medical health recommendation data to generate user feedback satisfaction detection data.

[0126] In an embodiment of the present invention, feedback information is collected from users through the first medical health recommendation system. User feedback information data includes two parts: user interaction data and user direct feedback data. User interaction data specifically includes the interaction behavior between the user and the recommendation system, such as clicks, browsing, collections, etc. User direct feedback data specifically includes the opinions, suggestions or comments directly expressed by the user through text, ratings, etc. Analyze the user's interactive behavior data and calculate the user's interaction intensity for different recommended contents. The user's preferences and content of concern can be identified through statistical methods or machine learning models, and then the user's interests and needs can be understood. Perform sentiment analysis on the user's direct feedback data to evaluate the user's emotional attitude towards the recommended content. Use natural language processing technology, such as sentiment dictionaries, machine learning models, etc. to analyze the emotional polarity (such as positive, negative, neutral) and emotional intensity in user feedback. Combine the user's interaction intensity data and direct feedback emotional data to calculate the health recommendation index. Methods such as weighted average and logistic regression can be used to comprehensively consider the user's behavior and emotional feedback to quantify the quality of the recommendation and user satisfaction. Evaluate the effect and user satisfaction of the first medical health recommendation data according to the health recommendation index. The test results reflect the user's overall reaction and evaluation of the recommendation system, helping to optimize and adjust the recommendation strategy.

[0127] As an example of the present invention, refer to Figure 4 As shown, in this example, step S4 includes:

[0128] Step S41: Containerizing the second medical health recommendation data to generate a health data containerization recommendation model;

[0129] Step S42: using the health data containerization recommendation model to access the second medical health recommendation data in real-time data stream, thereby generating a real-time medical recommendation data stream;

[0130] Step S43: constructing personalized recommendation decisions on the real-time medical recommendation data stream by implementing a decision engine to generate a medical health recommendation decision stream;

[0131] Step S44: Dynamically recommend terminal publishing of the medical and health recommendation decision flow, thereby generating a real-time interface for medical and health information recommendation.

[0132] In an embodiment of the present invention, the application, service and dependencies of the second medical health recommendation data are packaged into a container, such as a Docker container. Ensure that the containerized recommendation model can run independently, is highly portable, and has a consistent environment. Deploy the health data containerized recommendation model in a containerized environment to ensure that it can handle real-time recommendation requests, wherein a suitable containerization platform is selected, such as Docker, Kubernetes, etc. Select a suitable base image (such as python:3.8, ubuntu:20.04, etc.). Add the software and libraries required for installation in the Dockerfile, such as data processing libraries (pandas, numpy, etc.) and recommendation system libraries (scikit-learn, TensorFlow, etc.). Copy the processing scripts, model files, configuration files, etc. to the image. Use docker build -t <image-name> : <tag>. command to build a Docker image. Use the docker images command to view the generated image and verify it. Use docker run -d --name <container-name> <image-name> : <tag>Command to start the container. Configure environment variables, port mapping and data mounting (such as -p, -v parameters). Use docker logs <container-name>Command to monitor container logs. Use docker stats command to monitor container performance, such as CPU and memory usage. Design and implement data stream access mechanism to ensure that the second medical health recommendation data can be input into the recommendation model in a streaming manner, such as using data stream processing tools (such as Apache Kafka, Apache Flink) to access new data in real time. Transmit real-time data to the recommendation system in the container through the API interface. Design and implement a decision engine to make personalized recommendation decisions for users based on real-time medical recommendation data streams. The decision engine can include a rule engine, a machine learning model, or a hybrid method to generate personalized recommendation strategies based on the user's health needs, historical behaviors, and contextual information. Publish the medical health recommendation decision stream processed by the decision engine to a dynamic recommendation terminal, such as a web page, mobile application, or other user interface. Ensure that the recommendation interface can display personalized medical health recommendation information in a timely and accurate manner to improve user experience and satisfaction. More specifically, the process of generating personalized recommendation strategies based on the user's health needs, historical behaviors, and contextual information includes several key steps. First, collect the user's health needs, historical behavior data, and contextual information. Then, clean and preprocess these data to extract relevant features. Then, select a suitable recommendation algorithm, such as content-based recommendation, collaborative filtering, or hybrid recommendation, to generate a personalized recommendation strategy. When generating the strategy, consider the user's health goals, historical behavior, and real-time context information. Finally, implement the recommendation system, update the recommended content in real time, monitor system performance, collect user feedback, and make optimization adjustments based on the feedback.

[0133] In this specification, an information recommendation system for medical and health data mining is provided, which is used to execute the above-mentioned information recommendation method for medical and health data mining. The information recommendation system for medical and health data mining includes:

[0134] The knowledge graph construction module is used to obtain multi-source medical data; extract medical keywords from the multi-source medical data to obtain medical information keyword data; use the medical information keyword data to construct a medical knowledge graph for standard multi-source medical data to obtain a multi-source medical knowledge graph;

[0135] A shallow medical mining module is used to obtain user physiological health information data; perform abnormal physiological hazard assessment on the user's physiological health information data to generate user physiological abnormal hazard data; import the user's physiological abnormal hazard data into a multi-source medical knowledge graph to perform first medical health data mining, thereby generating first medical health recommendation data;

[0136] A deep medical mining module is used to collect recommendation feedback information from users based on the first medical health recommendation data to obtain user feedback information data; dynamically adjust the score recommendation of the first medical health recommendation data according to the user feedback information data, thereby generating second medical health recommendation data;

[0137] The decision delivery module is used to access the real-time data stream of the second medical and health recommendation data to generate a real-time medical recommendation data stream; construct personalized recommendation decisions for the real-time medical recommendation data stream to generate a medical and health recommendation decision stream; and dynamically recommend the terminal to publish the medical and health recommendation decision stream to generate a real-time interface for medical and health information recommendation.

[0138] The beneficial effect of the present invention is that by acquiring multi-source medical data and user physiological health information data, the system can integrate multiple data sources to obtain more comprehensive and accurate medical information. By constructing a knowledge graph for medical information keyword data, the system can organize medical knowledge into a structured form, providing more efficient and convenient knowledge acquisition and recommendation. Abnormal assessment of user physiological health information data can help identify potential physiological abnormality risks. Through dynamic adjustment of user feedback information data and score recommendations, the system can continuously optimize the recommendation results according to the user's personalized needs and feedback, and provide recommendation data that is more in line with the user's health status and preferences. By constructing personalized recommendation decisions for real-time medical recommendation data streams and generating a real-time interface for medical health information recommendations, the system can provide real-time, personalized medical health information recommendation services to help users obtain and apply relevant medical knowledge and health advice in a timely manner. Therefore, the present invention improves the accuracy and personalization of medical health data mining information recommendations through multi-source data acquisition, personalized recommendations, dynamic adjustments and real-time updates.

[0139] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0140] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein. < / tag> < / image-name> < / container-name> < / tag> < / image-name>

Claims

1. An information recommendation method applied to medical health data mining, characterized in that: The following steps are involved: Step S1: Acquire multi-source medical data; Perform data preprocessing on multi-source medical data to generate standard multi-source medical data; Extract medical keywords from multi-source medical data to obtain medical information keyword data; The medical information keyword data is used to construct a medical knowledge graph for standard multi-source medical data to obtain a multi-source medical knowledge graph; step S1 includes the following steps: Step S11: Acquire multi-source medical data; Step S12: performing data preprocessing on the multi-source medical data to generate standard multi-source medical data, wherein the data preprocessing includes data cleaning, data missing value filling and data standardization; Step S13: extracting medical keywords from standard multi-source medical data using natural language technology to obtain medical information keyword data; Step S14: constructing a medical knowledge graph for standard multi-source medical data using medical information keyword data to obtain a multi-source medical knowledge graph; Step S14 includes the following steps: Step S141: semantically analyzing the medical information keyword data to obtain medical information semantic relationship data; Step S142: performing semantic relationship recognition on the medical information keyword data through the medical information semantic relationship data to generate a medical information semantic relationship matrix; Step S143: extracting entity names from standard multi-source medical data based on the medical information semantic relationship matrix to obtain medical information entity naming data; Step S144: performing homonym disambiguation on the medical information entity naming data, thereby generating a standardized medical entity set; constructing a medical ontology based on the medical information keyword data and the standardized medical entity set, thereby generating a medical ontology; and performing entity association on the standardized medical entity set through the medical ontology, thereby generating a multi-source medical knowledge graph; Step S2: obtaining user physiological health information data; performing abnormal physiological hazard assessment on the user physiological health information data to generate user physiological abnormal hazard data; importing the user physiological abnormal hazard data into a multi-source medical knowledge graph to perform first medical health data mining, thereby generating first medical health recommendation data; Step S2 includes the following steps: Step S21: Acquire user physiological health information data; Step S22: identifying abnormal physiological health parameters of the user's physiological health information data to obtain abnormal physiological parameters of the user; Step S23: Performing a hazard assessment on the abnormal physiological parameters of the user to generate abnormal physiological hazard data of the user; Step S23 includes the following steps: Step S231: performing time series change analysis on the abnormal physiological parameters of the user to generate abnormal physiological change data of the user; performing fast Fourier transform based on the abnormal physiological change data of the user to generate a spectrum diagram of abnormal physiological change of the user; performing frequency domain analysis on the spectrum diagram of abnormal physiological change of the user to generate frequency domain data of abnormal physiological change of the user; Step S232: performing health stability assessment on the frequency domain data of abnormal physiological changes of the user to generate physiological health stability data of the user; Step S233: performing an abnormality over-limit analysis on the user's physiological abnormality parameters to generate user physiological abnormality over-limit data; performing an event-induced association analysis on the user's physiological abnormality over-limit data according to a multi-source medical knowledge graph to generate event-induced association data; Step S234: using the user's physiological health stability data to weight the event-induced correlation data for the harmfulness of the associated events, and obtaining the user's physiological abnormality harmfulness data; Step S24: importing the user's physiological abnormality hazard data into the multi-source medical knowledge graph to perform first medical health data mining, thereby generating first medical health recommendation data; Step S24 includes the following steps: Step S241: confirming the hazard impact factor of the user's physiological abnormality hazard data to obtain the physiological abnormality hazard impact factor; Step S242: importing physiological abnormality hazard influencing factors into the multi-source medical knowledge graph to perform hazard keyword search to obtain hazard keyword associated data; performing association classification on the multi-source medical knowledge graph according to the hazard keyword associated data to generate hazard overlap associated data and hazard non-overlap associated data; Step S243: performing mining priority division on the hazard overlap association data and the hazard non-overlap association data to generate association mining priority data; performing first medical health data mining on the user's physiological abnormality hazard data through the association mining priority data to generate first medical health recommendation data; Step S3: collecting recommendation feedback information of users based on the first medical health recommendation data to obtain user feedback information data; dynamically adjusting the score recommendation of the first medical health recommendation data according to the user feedback information data, thereby generating second medical health recommendation data; Step S3 includes the following steps: Step S31: collecting recommendation feedback information of users based on the first medical health recommendation data to obtain user feedback information data; performing user feedback satisfaction detection on the user feedback information data to generate user feedback satisfaction detection data; Step S32: recommending and scoring the first medical health recommendation data according to the user feedback satisfaction test data, and generating positive recommendation score data and negative recommendation score data; Step S33: hiding the related recommendations of the first medical health recommendation data according to the negative recommendation score data, thereby generating medical health negative recommendation optimization data; dynamically adjusting the recommended items of the first medical health recommendation data based on the positive recommendation score data, thereby generating medical health positive recommendation optimization data; Step S34: integrating the medical health negative recommendation optimization data and the medical health positive recommendation optimization data to generate second medical health recommendation data; Step S4: access the second medical and health recommendation data in real time to generate a real-time medical recommendation data stream; construct personalized recommendation decisions on the real-time medical recommendation data stream to generate a medical and health recommendation decision stream; dynamically publish the medical and health recommendation decision stream to a terminal to generate a real-time interface for medical and health information recommendation.

2. The information recommendation method for medical health data mining according to claim 1, characterized in that: Step S31 includes the following steps: Step S311: collecting recommendation feedback information of the user based on the first medical health recommendation data to obtain user feedback information data, wherein the user feedback information data includes user interaction data and user direct feedback data; Step S312: performing user interaction intensity analysis on the user interaction data to generate user interaction intensity data; performing feedback sentiment analysis on the user direct feedback data to generate user direct feedback sentiment data; Step S313: Perform recommendation scoring based on the user's direct feedback emotion data and user interaction intensity data to generate a health recommendation index; Step S314: Use the health recommendation index to perform user feedback satisfaction detection on the first medical health recommendation data to generate user feedback satisfaction detection data.

3. The information recommendation method for medical health data mining according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: Containerizing the second medical health recommendation data to generate a health data containerization recommendation model; Step S42: using the health data containerization recommendation model to access the second medical health recommendation data in real-time data stream, thereby generating a real-time medical recommendation data stream; Step S43: constructing personalized recommendation decisions on the real-time medical recommendation data stream by implementing a decision engine to generate a medical health recommendation decision stream; Step S44: Dynamically recommend terminal publishing of the medical and health recommendation decision flow, thereby generating a real-time interface for medical and health information recommendation.

4. An information recommendation system applied to medical and health data mining, characterized in that: Used to execute the information recommendation method for medical health data mining as claimed in claim 1, the information recommendation system for medical health data mining comprises: The knowledge graph construction module is used to obtain multi-source medical data; extract medical keywords from the multi-source medical data to obtain medical information keyword data; use the medical information keyword data to construct a medical knowledge graph for standard multi-source medical data to obtain a multi-source medical knowledge graph; A shallow medical mining module is used to obtain user physiological health information data; perform abnormal physiological hazard assessment on the user's physiological health information data to generate user physiological abnormal hazard data; import the user's physiological abnormal hazard data into a multi-source medical knowledge graph to perform first medical health data mining, thereby generating first medical health recommendation data; A deep medical mining module is used to collect recommendation feedback information from users based on the first medical health recommendation data to obtain user feedback information data; dynamically adjust the score recommendation of the first medical health recommendation data according to the user feedback information data, thereby generating second medical health recommendation data; The decision delivery module is used to access the real-time data stream of the second medical and health recommendation data to generate a real-time medical recommendation data stream; construct personalized recommendation decisions for the real-time medical recommendation data stream to generate a medical and health recommendation decision stream; and dynamically recommend the terminal to publish the medical and health recommendation decision stream to generate a real-time interface for medical and health information recommendation.

Images