An Internet online diagnosis and treatment data processing system based on big data model

Abstract

The present invention discloses an Internet online diagnosis and treatment data processing system based on a big data model, which relates to the field of medical technology. The association rule analysis module uses weighted lifting to analyze the association between disease and drug treatment, which can accurately identify the relatively good matching relationship between disease and drug, and improve the pertinence and accuracy of drug recommendation. The online recommendation module uses candidate drug pairs, and combines the patient's real-time treatment situation, and dynamically analyzes the recovery effect of drugs in different time periods based on dynamic conditional probability. This mechanism not only ensures that the recommended drugs are highly matched with the patient's current condition, but also can adjust the drug selection according to the patient's feedback data. The system effectively avoids the problem of inaccurate recommendations caused by data sparsity or insufficient data in traditional systems by weighted processing of historical data and optimizing it in combination with data sparsity.

Description

Technical Field

[0001] The present invention relates to the field of medical technology, and in particular to an Internet online diagnosis and treatment data processing system based on a big data model. Background Art

[0002] In the field of modern medicine and health management, Internet online diagnosis and treatment, as an important part of health informatization, has gradually been widely used around the world. With the continuous development of medical technology, especially the maturity of big data and artificial intelligence technology, Internet online diagnosis and treatment has gradually become intelligent and personalized, providing patients with convenient online health consultation and treatment suggestions. In this large field, Internet online diagnosis and treatment data processing, as one of its core technologies, is gradually changing the traditional medical model, making medical resources more efficiently allocated and used, especially playing an important role in data analysis and patient treatment plan recommendation.

[0003] At present, although many online diagnosis and treatment systems based on the Internet can realize certain disease diagnosis and drug recommendation functions, they often face some significant problems in practical applications. One of the more prominent problems is data sparsity. In real life, patients' health data is often incomplete, especially in the fields of chronic diseases and rare diseases. There is often a lack of sufficient case data for effective model training, which makes it difficult to accurately discover the association rules between many drugs and diseases, which in turn affects the effect of drug recommendations. In addition, many existing systems lack sufficient dynamic feedback mechanisms, making it difficult to make real-time adjustments based on the patient's recovery. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides an Internet online diagnosis and treatment data processing system based on a big data model, which solves the problems in the above-mentioned background technology.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: an Internet online diagnosis and treatment data processing system based on a big data model, comprising a data collection module, an association rule analysis module, an online recommendation module, an error feedback module and an optimization module;

[0006] The data collection module is used to collect relevant historical data of patients to be diagnosed and treated by using big data technology, and obtain the disease of the current patient to be diagnosed and treated through the online diagnosis and treatment process on the Internet. Based on big data technology, relevant historical data of patients with the same disease as the patient to be diagnosed and treated are collected, and a database is generated after data preprocessing;

[0007] The association rule analysis module is used to build a big data model based on the database, analyze the association rules between the corresponding diseases and various types of drug treatments based on the big data model, so as to build a weighted lift degree Jtd, and determine the candidate pairs based on the weighted lift degree Jtd value;

[0008] The online recommendation module is used to dynamically analyze the recovery status of various types of drugs in the candidate pairs in various time periods based on the candidate pairs and the relevant historical data of patients with the same disease as the patient to be diagnosed in the database, so as to construct the dynamic conditional probability Dtgz. Through the dynamic conditional probability Dtgz, the corresponding drugs are given priority recommendation.

[0009] The error feedback module is used to recommend the drugs that are preferentially recommended in the online recommendation module to the patient to be diagnosed, and monitor the recovery status of the patient to be diagnosed in real time when using the corresponding drugs to obtain the time when the patient to be diagnosed fully recovers, and determine the actual time difference Sc according to the time when the patient to be diagnosed fully recovers;

[0010] The optimization module is used to optimize the online treatment plan based on the actual time difference Sc.

[0011] Preferably, the data collection module includes a first collection unit, a second collection unit and a processing unit;

[0012] The No. 1 collection unit is used to utilize big data technology to collect relevant historical data of patients to be diagnosed, wherein the relevant historical data of patients to be diagnosed include the medical records and treatment plans of patients to be diagnosed within a historical period; and through the Internet online diagnosis and treatment process, establish a communication relationship with the patients to be diagnosed, and combine the relevant historical data of the patients to be diagnosed to obtain the current diseases of the patients to be diagnosed and the current relevant pathological data.

[0013] The second acquisition unit is used to collect relevant historical data of patients with the same disease as the patient to be diagnosed based on big data technology, and to treat patients with the same disease as the patient to be diagnosed as non-target patients, where the relevant historical data of non-target patients includes the frequency of occurrence of disease X , the frequency of occurrence of the corresponding type of drug Y in the corresponding item set and the case where disease X and the corresponding type of drug Y in the corresponding item set appear at the same time; where the item set refers to the various types of drugs adopted for disease X;

[0014] The processing unit is used to perform data preprocessing on the relevant historical data of the patients to be diagnosed and the relevant historical data of non-target patients to remove noise, identify and fill missing values ​​and abnormal values, and generate a database after dimensionless processing.

[0015] Preferably, the association rule analysis module includes a sparse data processing unit and an association analysis unit;

[0016] The sparse data processing unit is used to analyze the appearance status of each item set in the database according to the relevant historical data of non-target patients to calculate and obtain the weighted support Jzcd. The weighted support Jzcd is obtained by the following formula:

[0017] ;

[0018] In the formula, Reflects the frequency of co-occurrence of disease X and corresponding type of drug Y in the i-th item set, n represents the total number of item sets in the database, i=1, 2, 3, ..., n, represents the weight value of the i-th item set, represents the indicator function, which takes the value of 1 if the disease X and the corresponding type of drug Y in the i-th item set appear at the same time, otherwise it takes the value of 0;

[0019] Among them, the weight is adjusted according to the sparsity of each item set in the database to obtain the weight value of the corresponding item set , where the weight value of the i-th item set is Take the example to calculate the weight value of the i-th item set. :

[0020] ;

[0021] In the formula, Reflects the frequency of co-occurrence of disease X and corresponding type of drug Y in the i-th item set.

[0022] Preferably, the association analysis unit is used to analyze the association rules between the corresponding disease and each type of drug treatment based on the relevant historical data of the non-target patients and in combination with the weighted support Jzcd, so as to calculate the weighted lift Jtd. The weighted lift Jtd is specifically obtained by the following formula:

[0023] ;

[0024] In the formula, Reflects the frequency of simultaneous occurrence of disease X and corresponding type of drug Y in the i-th item set, represents the frequency of disease X, Represents the frequency of occurrence of the corresponding type of drug Y in the i-th item set.

[0025] Preferably, according to the method of obtaining the weighted lift Jtd, the frequencies of the disease X and the corresponding type of drug Y in each set appearing at the same time are calculated and marked as the weighted lift Jtd of each set, and the association threshold K is pre-set, and compared and analyzed with the weighted lift Jtd of each set to determine the candidate pairs of patients to be diagnosed and treated, as follows:

[0026] If the weighted lift Jtd of each item set exceeds the association threshold K, it indicates that there is an association rule between the corresponding disease in the current item set and the corresponding type of drug treatment. At this time, the corresponding disease in the current item set and the corresponding type of drug treatment are marked as candidate pairs;

[0027] If the weighted lift Jtd of each item set does not exceed the association threshold K, it indicates that there is no association rule between the corresponding disease in the current item set and the corresponding type of drug treatment. At this time, the corresponding disease in the current item set and the corresponding type of drug treatment are not marked as candidate pairs.

[0028] Preferably, the online recommendation module includes a dynamic analysis unit, a data constraint processing unit and a priority recommendation unit;

[0029] The dynamic analysis unit is used to extract the corresponding diseases and corresponding types of drug treatments involved in the candidate pairs from the relevant historical data of non-target patients according to the candidate pairs, and the relevant historical data of each period in the process of mutual matching and patient adoption, and combine the current relevant pathological data of the patient to be diagnosed and treated to dynamically analyze the recovery of each type of drug treatment in the candidate pairs in each period, and construct the dynamic conditional probability Dtgz. The dynamic conditional probability Dtgz is obtained by the following formula:

[0030] ;

[0031] In the formula, represents the joint probability of the non-target patient's recovery event, treatment plan, and the current severity of the patient's illness occurring simultaneously at time j; Indicates that at time j, non-target patients adopt the treatment plan And it has the probability of the severity of the current illness of the patient to be diagnosed; represents the recovery event of non-target patients at time j; represents the g-th treatment plan adopted by the non-target patient at time j; It indicates the current severity of the patient's condition at time j; j represents time;

[0032] Among them, at time j, the recovery event of non-target patients Includes complete recovery, partial recovery, and no recovery;

[0033] At time j, the gth treatment plan adopted by the non-target patient Refers to the original drug dosage taken when taking each type of drug for the corresponding disease and the time of taking;

[0034] At time j, the current severity of the patient's condition The specific acquisition method is as follows:

[0035] ;

[0036] In the formula, Indicates the duration of the disease in the relevant pathological data, represents the frequency of disease attacks in the relevant pathological data, and Disease duration and frequency of disease onset The weight value of <1,0< <1, and The specific value is set by the user according to the situation.

[0037] Preferably, the data constraint processing unit is used to constrain the output value of the dynamic conditional probability Dtgz using a Softmax function to ensure that the rehabilitation events of multiple groups of non-target patients are converted into probabilities, and the sum of the probabilities of the rehabilitation events of the multiple groups of non-target patients is constrained to 1;

[0038] The priority recommendation unit is used to obtain multiple groups of dynamic conditional probabilities Dtgz at different time points and different recovery events according to the method of obtaining the dynamic conditional probability Dtgz in the dynamic analysis unit, and extract the dynamic conditional probability Dtgz of the earliest complete recovery from them, so as to mark the candidate pair corresponding to the earliest complete recovery as a recommended pair, and determine the type of medicine in the recommended pair, so as to make priority recommendations to the patients to be diagnosed and treated.

[0039] Preferably, the error feedback module includes a collection unit and a time difference comparison unit;

[0040] The collection unit is used to recommend the drugs that are prioritized in the online recommendation module to the patient to be diagnosed, and monitor the recovery status of the patient to be diagnosed in real time when using the corresponding drugs to obtain the time when the patient to be diagnosed is fully recovered, and mark the time when the patient to be diagnosed is fully recovered as the actual time. .

[0041] Preferably, the time difference comparison unit is used to determine the predicted time according to the dynamic conditional probability Dtgz of the earliest complete recovery extracted from the priority recommendation unit. , according to the predicted time , combined with the actual time obtained in the acquisition unit , determine the actual time difference Sc, the specific acquisition method is: ;

[0042] An error threshold W is set in advance, and the actual time difference Sc is compared and analyzed with the error threshold W to determine whether the treatment plan adopted by the current patient to be treated needs to be optimized. If the actual time difference Sc exceeds the error threshold W, it indicates that the current error is abnormal, and the optimization mechanism will be activated at this time; if the actual time difference Sc does not exceed the error threshold W, it indicates that the current error is not abnormal, and the optimization mechanism will not be activated at this time.

[0043] Preferably, the optimization module is used to start and execute the optimization mechanism to adjust the original drug dosage taken in the treatment plan according to the actual time difference Sc , in order to optimize the online treatment program, the specific adjustments are as follows:

[0044] ;

[0045] In the formula, represents the adjusted drug dose, Represents the learning rate, which controls the magnitude of each adjustment; Indicates the prediction time , Indicates the actual time difference, It represents the ratio of the actual time difference to the predicted time.

[0046] The present invention provides an Internet online diagnosis and treatment data processing system based on a big data model, which has the following beneficial effects:

[0047] (1) Through the data collection module, the system can accurately obtain the historical health data of the patient to be diagnosed and treated, and combine it with the data of other patients with similar diseases to build a database based on the big data model. This process ensures the scientificity and personalization of the diagnosis and treatment plan recommendation, and tailors the treatment plan for each patient to ensure the effectiveness of the treatment. Improve the accuracy of treatment recommendations: The association rule analysis module uses weighted lifting to analyze the relationship between disease and drug treatment. Through in-depth mining of these association rules, it can accurately identify the relatively good matching relationship between disease and drug, improve the pertinence and accuracy of drug recommendations, and dynamically track and optimize with real-time feedback: The online recommendation module uses candidate drug pairs and combines the patient's real-time treatment situation to dynamically analyze the recovery effect of drugs in different time periods based on dynamic conditional probability. This mechanism not only ensures that the recommended drugs are highly matched with the patient's current condition, but also can adjust the drug selection according to the patient's feedback data to achieve real-time optimization and avoid the situation of inadaptability or poor effect during the treatment process. Provide optimization solutions for data sparsity issues: The system effectively avoids the problem of inaccurate recommendations caused by data sparsity or insufficient data in traditional systems by weighted processing of historical data and optimizing it in combination with data sparsity. The combination of weighted lift and dynamic conditional probability ensures that the system can discover potentially valuable relationships between drugs and diseases even when data is sparse. Accurate recovery time prediction and plan optimization: Through the error feedback module, the system can monitor the patient's recovery status after using the recommended drugs in real time, and quantify the difference between the actual recovery time and the predicted recovery time. Based on this feedback, the system can further optimize the treatment plan, reduce treatment errors, and improve the accuracy and efficiency of treatment. Through multiple iterative optimizations, more personalized and accurate treatment plan recommendations can be achieved.

[0048] (2) The system effectively addresses data sparsity issues and optimizes treatment plans through dynamic conditional probability and feedback mechanisms. First, the system uses weighted lift to analyze the degree of association between different drugs and diseases. For sparse data, it uses methods such as inverse document frequency for weighting to ensure that the potential value of sparse data can be fully tapped. Secondly, by constructing dynamic conditional probabilities based on historical data and feedback data, the system can dynamically adjust recommended plans based on the patient's recovery in different time periods and provide relatively accurate personalized treatment recommendations. Finally, through the error feedback mechanism, the system can track the patient's recovery in real time, adjust drug recommendations based on actual time differences, and optimize treatment plans to maximize treatment outcomes. Through this series of optimizations, the system can effectively overcome data sparsity issues and achieve more accurate and personalized online diagnosis and treatment services.

[0049] (3) In the association rule analysis module, the core function of the sparse data processing unit is to analyze and process the characteristics of sparse data in the database, and use these sparse data to reveal possible implicit associations. These sparse data items (i.e., item sets with low frequency of occurrence) often contain more unique and valuable association information. By assigning higher weights to sparse data items, some potential association patterns that are ignored by traditional data analysis methods can be discovered, thereby improving the diversity and adaptability of treatment plans, especially for a few special diseases.

[0050] (4) In the dynamic analysis unit, the system dynamically analyzes the diseases and drug treatments in the candidate pairs, combines historical data and the current pathological data of the patient to be treated, and constructs dynamic conditional probabilities. In this way, the system can dynamically analyze the treatment effects in different time periods and evaluate the recovery of patients with different drugs, thereby providing patients with personalized and timely treatment recommendations. In the data constraint processing unit, the system uses the Softmax function to constrain the dynamic conditional probability to ensure that the sum of the probabilities of multiple groups of recovery events is 1. This process ensures that the probability value of the treatment recommendation is within a reasonable range and avoids the situation where the sum of the probabilities exceeds 1. In the priority recommendation unit, the system prioritizes treatment plans based on dynamic conditional probabilities, ensuring that those drugs or treatment plans that show a higher probability of recovery in multiple dynamic analyses are given priority. By continuously iteratively analyzing the effects of different treatment plans at various time points, the recommendation system can be optimized according to the specific situation of the patient, thereby improving the success rate of treatment and the speed of patient recovery. Personalized treatment recommendations: By dynamically analyzing the severity of the patient's condition and historical data, combined with the effects of multiple drug treatment options, the system can provide personalized treatment recommendations for each patient. This dynamic recommendation system based on the patient's individualized data and historical patterns can further refine treatment, thereby improving treatment efficiency and the patient's recovery speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a block diagram of an Internet online diagnosis and treatment data processing system based on a big data model of the present invention. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] Example 1

[0054] See also Figure 1, the present invention provides an Internet online diagnosis and treatment data processing system based on a big data model, including a data collection module, an association rule analysis module, an online recommendation module, an error feedback module and an optimization module;

[0055] The data collection module is used to collect relevant historical data of patients to be diagnosed and treated by using big data technology, and obtain the disease of the current patient to be diagnosed and treated through the online diagnosis and treatment process on the Internet. Based on big data technology, relevant historical data of patients with the same disease as the patient to be diagnosed and treated are collected, and a database is generated after data preprocessing;

[0056] The association rule analysis module is used to build a big data model based on the database. Based on the big data model, the relevant historical data of patients with the same disease as the patient to be diagnosed in the database are analyzed to analyze the association rules between the corresponding disease and various types of drug treatments to build a weighted lift degree Jtd. Based on the weighted lift degree Jtd value, the candidate pairs are determined;

[0057] Use deep learning technology to build an initial model, and train and test the initial model with relevant data in the database. Use the trained initial model as a state recognition model, obtain feature information in the state recognition model, and use the obtained feature information to train and test the state recognition model. Finally, use the trained state recognition model as a big data model.

[0058] The online recommendation module is used to dynamically analyze the recovery status of various types of drugs in the candidate pairs in various time periods based on the candidate pairs and the relevant historical data of patients with the same disease as the patient to be diagnosed in the database, so as to construct the dynamic conditional probability Dtgz. Through the dynamic conditional probability Dtgz, the corresponding drugs are given priority recommendation.

[0059] The error feedback module is used to recommend the drugs that are preferentially recommended in the online recommendation module to the patient to be diagnosed, and monitor the recovery status of the patient to be diagnosed in real time when using the corresponding drugs to obtain the time when the patient to be diagnosed fully recovers, and determine the actual time difference Sc according to the time when the patient to be diagnosed fully recovers;

[0060] The optimization module is used to optimize the online treatment plan based on the actual time difference Sc.

[0061] During the operation of this system, the personalization and accuracy of the treatment plan are improved: the data collection module uses big data technology to comprehensively collect the patient's historical health data, and combined with the Internet online diagnosis and treatment process, the system can obtain the disease information of the patient to be diagnosed in real time. This data is further preprocessed and a database is generated to provide accurate basic information for subsequent analysis and recommendation. Compared with traditional diagnosis and treatment plans, the system can not only consider the patient's disease history, but also make personalized recommendations based on the historical data of similar patients, ensuring that the treatment plan is more accurate and meets the individual needs of the patient. Overcome the problem of data sparsity and improve the accuracy of drug recommendations: through the association rule analysis module, the system can analyze the association rules between diseases and various types of drug treatments based on the patient's historical data in the database, and construct a weighted lift Jtd. The weighted lift can effectively solve the problem of data sparsity. By calculating the strength of the association between drugs and diseases in the patient group, it helps to identify drugs related to the patient's disease, thereby improving the accuracy of drug recommendations. Compared with traditional methods, the present invention uses big data analysis technology to better deal with the challenges of sparse data and ensure the scientific nature of the treatment plan. Dynamically adjust the recommended plan to improve the response speed of treatment effect: The online recommendation module combines candidate pairs and uses the dynamic conditional probability Dtgz model to dynamically analyze the patient's recovery during different treatment periods. By dynamically adjusting drug recommendations, the system can optimize the treatment strategy in real time. This real-time and dynamic adjustment mechanism greatly improves the flexibility of the treatment process, allowing patients to obtain treatment plans that are relatively suitable for their individual characteristics more quickly. Continuously optimize the treatment plan through the feedback mechanism: The error feedback module can optimize the difference between the actual recovery time and the predicted time (actual time difference Sc) by monitoring the patient's recovery during drug use in real time. This feedback mechanism can not only improve the real-time and accuracy of drug recommendations, but also further adjust the subsequent treatment plan according to the patient's actual recovery, thereby shortening the patient's recovery time as much as possible. Comprehensively improve the efficiency of diagnosis and treatment and optimize patient experience: Comprehensively utilizing data collection, association rule analysis, dynamic recommendation, error feedback and optimization modules, the system can respond to patient feedback in real time while ensuring the scientific nature of the treatment plan, continuously optimize the recommendation results, and comprehensively improve the patient's treatment experience. This not only improves the patient's recovery speed, but also optimizes the plan according to the patient's individual differences and real-time data, reduces unnecessary treatment steps and drug use, and optimizes resource allocation.

[0062] Example 2

[0063] Please refer to Figure 1 ,Specifically: the data collection module includes a No. 1 collection unit, a No. 2 collection unit and a processing unit;

[0064] The No. 1 collection unit is used to utilize big data technology to collect relevant historical data of patients to be diagnosed, wherein the relevant historical data of patients to be diagnosed include the medical records and treatment plans of patients to be diagnosed within a historical period; and through the Internet online diagnosis and treatment process, establish a communication relationship with the patients to be diagnosed, and combine the relevant historical data of the patients to be diagnosed to obtain the current diseases of the patients to be diagnosed and the current relevant pathological data.

[0065] The second acquisition unit is used to collect relevant historical data of patients with the same disease as the patient to be diagnosed based on big data technology, and to treat patients with the same disease as the patient to be diagnosed as non-target patients, where the relevant historical data of non-target patients includes the frequency of occurrence of disease X , the frequency of occurrence of the corresponding type of drug Y in the corresponding item set and the case where disease X and the corresponding type of drug Y in the corresponding item set appear at the same time; where the item set refers to the various types of drugs adopted for disease X;

[0066] The processing unit is used to perform data preprocessing on the relevant historical data of the patients to be diagnosed and the relevant historical data of non-target patients to remove noise, identify and fill missing values ​​and abnormal values, and generate a database after dimensionless processing.

[0067] In this embodiment, multi-dimensional data collection comprehensively improves the quality of diagnosis and treatment data: through the No. 1 collection unit, the system can not only collect historical cases and treatment plans of patients to be diagnosed, but also establish communication relationships with patients through the Internet in real time, collect current disease information and related pathological data. This multi-faceted data collection method ensures that the system can obtain multi-dimensional information of patients in historical periods and current states, and provides sufficient data support for subsequent treatment plan recommendations. Rich comparative data sources improve the accuracy of disease and drug association analysis: The No. 2 collection unit collects patient data with the same disease as the patient to be diagnosed through big data technology, and regards these patients as non-target patients for comparative analysis. By collecting the frequency of disease occurrence, drug use frequency, and the simultaneous occurrence of disease and drug related to non-target patients, the system can construct more representative and extensive disease-drug association data, providing a reliable data basis for subsequent association rule mining. Data preprocessing and cleaning improve the accuracy and completeness of data: The processing unit removes noise, fills and identifies missing values ​​and outliers based on the collected data, ensuring that the data input to the system is more accurate, complete and consistent. In addition, through dimensionless processing, the system can eliminate the interference caused by different data scales and ensure that various types of data can be effectively analyzed under the same framework. Build a high-quality database to enhance the reliability and effectiveness of subsequent data analysis: Through the above-mentioned preprocessing work, the generated database contains comprehensive, clean and standardized data, which provides high-quality data support for the work of subsequent modules such as association rule analysis module and online recommendation module. The optimized database not only improves the efficiency of data mining, but also provides accurate basic data for subsequent disease-drug association analysis and dynamic recommendation based on big data model. Solve the problem of data sparsity and enhance the learning ability of the model: Through the collection of historical data and real-time pathological data of a large number of patients, the system can effectively solve the problem of data sparsity, making the recommendation of treatment plans more comprehensive and applicable to different patients, which provides a reliable reference for the subsequent construction of accurate treatment plans, making the recommendation results more personalized and targeted. Through the realization of the above beneficial effects, the data collection module of the present invention lays a solid foundation for subsequent drug recommendations, treatment plan optimization and other processes, thereby improving the accuracy, real-time and personalization level of the overall Internet online diagnosis and treatment system.

[0068] Example 3

[0069] Please refer to Figure 1 ,Specifically: the association rule analysis module includes a sparse data processing unit and an ,association analysis unit;

[0070] The sparse data processing unit is used to analyze the appearance status of each item set in the database according to the relevant historical data of non-target patients to calculate and obtain the weighted support Jzcd. The weighted support Jzcd is obtained by the following formula:

[0071] ;

[0072] In the formula, Reflects the frequency of co-occurrence of disease X and corresponding type of drug Y in the i-th item set, n represents the total number of item sets in the database, i=1, 2, 3, ..., n, represents the weight value of the i-th item set, represents the indicator function, which takes the value of 1 if the disease X and the corresponding type of drug Y in the i-th item set appear at the same time, otherwise it takes the value of 0;

[0073] Among them, the weight is adjusted according to the sparsity of each item set in the database to obtain the weight value of the corresponding item set , where the weight value of the i-th item set is Take the example to calculate the weight value of the i-th item set. : The weighted support Jzcd reflects the frequency of co-occurrence in the corresponding item set;

[0074] ;

[0075] The weight of the i-th item set It is used to measure the sparsity of an itemset, that is, the frequency of the itemset appearing in the entire database. If the itemset rarely appears, it is considered sparse. The value will be larger, which means that the item set is relatively important; A higher value means that the item set is relatively sparse, that is, it appears less frequently in the data set. Because it appears less frequently, it is somewhat neglected in the existing technology, so its information value cannot be underestimated, so it is given a greater weight. A lower value: indicates that the item set is more common, that is, it appears more frequently in the database. Because it appears frequently, its information value is relatively low, so its weight will be smaller.

[0076] Why sparse data items are more important: Sparse data items (i.e., itemsets with low frequency) usually contain more potential and unique association information. In many data analyses, sparse data may represent more special patterns that are often underestimated in larger data sets. Weighted support can give more attention to less frequent patterns by weighting the weight of each item set.

[0077] In this embodiment, the challenge of sparse data is addressed: In many practical applications, many important associations in the data may exist in itemsets with low frequency of occurrence, that is, in sparse data. These sparse itemsets contain a large amount of potential important information, but in traditional data analysis methods, these itemsets may be ignored or underestimated due to their low frequency of occurrence. By weighting the sparse data items, their importance in the analysis can be enhanced, ensuring that the sparse data is fully considered and potential valuable association rules are mined. Association rule optimization based on weighted support: By calculating the weighted support, the strength of the association between the disease and the drug can be effectively measured. The weighted support reflects the frequent occurrence of a disease and a specific drug combination in the data set, and adjusts the weight of the item set according to the sparsity, so that those itemsets with low frequency of occurrence but important information can obtain higher weights, so that the system can more accurately identify those drug treatment combinations that are uncommon in most patients but may be very effective. Improve the accuracy and effectiveness of data analysis: In traditional association rule mining, frequent itemsets and common patterns usually dominate, while sparse itemsets may be ignored. Strengthen personalized recommendation capabilities: In the Internet online diagnosis and treatment system, sparse data processing enables the system to accurately identify the relationship with other similar patients based on each patient's personalized medical history, thereby recommending a drug combination that is relatively in line with the treatment needs of the patient to be diagnosed. Improve treatment effects and patient satisfaction by optimizing drug recommendations and treatment options. Avoid over-reliance on common data: By focusing on sparse data items, over-reliance on common data in traditional association rule mining is avoided, ensuring that the system can discover more uncommon but potentially effective relationships between drugs and diseases. This method not only enhances the system's predictive ability, but also improves the model's ability to identify special cases to a large extent.

[0078] Example 4

[0079] Please refer to Figure 1 Specifically: the association analysis unit is used to analyze the association rules between the corresponding diseases and various types of drug treatments based on the relevant historical data of non-target patients and combined with the weighted support Jzcd to calculate the weighted lift Jtd. The weighted lift Jtd is specifically obtained by the following formula:

[0080] ;

[0081] In the formula, represents the frequency of disease X, represents the frequency of occurrence of the corresponding type of drug Y in the i-th item set;

[0082] The greater the lift, the stronger the rule.

[0083] According to the method of obtaining the weighted lift Jtd, the frequencies of the co-occurrence of the disease X and the corresponding type of drug Y in each set are calculated and marked as the weighted lift Jtd of each set. The association threshold K is pre-set and compared with the weighted lift Jtd of each set to determine the candidate pairs of patients to be diagnosed and treated, as follows:

[0084] If the weighted lift Jtd of each item set exceeds the association threshold K, it indicates that there is an association rule between the corresponding disease in the current item set and the corresponding type of drug treatment. At this time, the corresponding disease in the current item set and the corresponding type of drug treatment are marked as candidate pairs;

[0085] If the weighted lift Jtd of each item set does not exceed the association threshold K, it indicates that there is no association rule between the corresponding disease in the current item set and the corresponding type of drug treatment. At this time, the corresponding disease in the current item set and the corresponding type of drug treatment are not marked as candidate pairs.

[0086] In this embodiment, the association analysis unit uses weighted support Jzcd and weighted lift Jtd to quantify the relationship between disease X and various types of drugs Y. By calculating the frequency of occurrence of each item set (combination of disease and drug) in historical data and the lift of the combination, the system can identify which diseases and drugs have significant statistical associations. The system sets an association threshold K to screen disease-drug pairs with significant associations. This threshold-based screening mechanism enables the system to effectively extract useful treatment information from a large amount of historical data and avoid overly redundant and ineffective treatment recommendations. By marking candidate pairs of strong association rules, the system can recommend drug combinations that have been verified by historical data and are strongly related to the disease based on the disease information of a specific patient, which can not only reduce the blindness of treatment recommendations, but also improve the personalization of treatment plans. Especially in sparse data scenarios, the potential association between drugs and diseases is often ignored by traditional methods, and these implicit associations can be discovered through the calculation of weighted lift, further improving the accuracy of treatment plans. Through the dynamic calculation and threshold setting of the weighted lift Jtd, the system can quickly filter out irrelevant treatment information and focus on analyzing potential disease-drug combinations. This mechanism further improves the system's computational efficiency, allowing online diagnosis and treatment to respond quickly, providing timely and accurate treatment plans, reducing patient waiting time, and improving the efficiency of the diagnosis and treatment process. Through the above method, the association rule analysis module significantly improves the intelligence level of the online diagnosis and treatment system, enabling the system to extract valuable and accurate association rules from massive medical data. This mechanism can not only improve the accuracy of the correlation analysis between diseases and drug treatments, but also provide personalized treatment recommendations in real time based on the patient's specific situation, ultimately optimizing the diagnosis and treatment effect and improving the patient's health recovery speed.

[0087] Example 5

[0088] Please refer to Figure 1 ,Specifically: the online recommendation module includes a dynamic analysis unit, a ,data constraint processing unit and a priority recommendation unit;

[0089] The dynamic analysis unit is used to extract the corresponding diseases and corresponding types of drug treatments involved in the candidate pairs from the relevant historical data of non-target patients according to the candidate pairs, and the relevant historical data of each period in the process of mutual matching and patient adoption, and combine the current relevant pathological data of the patient to be diagnosed and treated to dynamically analyze the recovery of each type of drug treatment in the candidate pairs in each period, and construct the dynamic conditional probability Dtgz. The dynamic conditional probability Dtgz is obtained by the following formula:

[0090] ;

[0091] In the formula, represents the joint probability of the non-target patient's recovery event, treatment plan, and the current severity of the patient's illness occurring simultaneously at time j; Indicates that at time j, non-target patients adopt the treatment plan And it has the probability of the severity of the current illness of the patient to be diagnosed; represents the recovery event of non-target patients at time j; represents the g-th treatment plan adopted by the non-target patient at time j; It indicates the current severity of the patient's condition at time j; j represents time;

[0092] Among them, at time j, the recovery event of non-target patients Includes complete recovery, partial recovery, and no recovery;

[0093] At time j, the gth treatment plan adopted by the non-target patient Refers to the original drug dosage taken when taking each type of drug for the corresponding disease and the time of taking;

[0094] At time j, the current severity of the patient's condition The specific acquisition method is as follows:

[0095] ;

[0096] In the formula, Indicates the duration of the disease in the relevant pathological data, represents the frequency of disease attacks in the relevant pathological data, and Disease duration and frequency of disease onset The weight value of <1,0< <1, and The specific value is set by the user according to the situation.

[0097] The data constraint processing unit is used to constrain the output value of the dynamic conditional probability Dtgz using the Softmax function to ensure that the recovery events of multiple groups of non-target patients are converted into probabilities, and the sum of the probabilities of the recovery events of the multiple groups of non-target patients is constrained to 1;

[0098] The priority recommendation unit is used to obtain multiple groups of dynamic conditional probabilities Dtgz at different time points and different recovery events according to the method of obtaining the dynamic conditional probability Dtgz in the dynamic analysis unit, and extract the dynamic conditional probability Dtgz of the earliest complete recovery from them, so as to mark the candidate pair corresponding to the earliest complete recovery as a recommended pair, and determine the type of medicine in the recommended pair, so as to make priority recommendations to the patients to be diagnosed and treated.

[0099] Using the Softmax function to deal with multi-category problems does require some adjustments or optimizations to the dynamic conditional probability formula to adapt it to the context of multi-category outputs, where the Softmax function is typically used for multi-classification problems.

[0100] In this embodiment, based on the candidate pairs, matching data of related diseases and drug treatments are extracted from the historical data of non-target patients. Combined with the current morbid data of the patient to be diagnosed, the effect of drug treatment at different time points is dynamically analyzed, and the dynamic conditional probability Dtgy is calculated. The dynamic conditional probability takes into account the time factor, the severity of the disease and the recovery status of the patient during the treatment process. The dynamic conditional probability is constrained by the Softmax function to ensure that the patient's recovery events are converted into probability values ​​in each time period, and the sum of the probabilities of all events is 1, so that the priority of each treatment plan has reasonable normalization and avoids unreasonable probability distribution. According to the calculated dynamic conditional probability, combined with the actual severity of the patient's disease and the treatment plan, a relatively suitable drug treatment plan is recommended first, and an appropriate priority is assigned to each plan to ensure that relatively suitable treatment can be provided at different time nodes. Accurate treatment plan recommendation: The dynamic analysis unit can accurately simulate the effects of different treatment plans in each time period by combining the current morbid data of the patient to be diagnosed and the treatment effect of the non-target patient, which makes the recommended treatment plan more personalized and can more effectively respond to the patient's real-time condition changes. Efficient drug selection: The dynamic conditional probability is constrained by the Softmax function to ensure that the probability distribution of different treatment plans is reasonable and consistent. This constraint avoids the problem of inconsistent probabilities or inaccurate predictions, and ensures that the sum of the probabilities of multiple groups of candidate treatment plans is 1, so that drugs with possible cure can be given priority when recommending. Dynamic feedback and adaptability: The online recommendation module combines the historical data of non-target patients and the current pathological information of patients to be diagnosed and treated. It not only considers the frequency of disease occurrence, but also combines multiple dimensions such as the duration of the disease and the frequency of onset for comprehensive evaluation. It can flexibly adapt to the needs of different patients and provide dynamic treatment plans. This treatment recommendation mechanism based on real-time patient data adjustment can significantly improve the treatment effect and shorten the patient's recovery time. Optimize treatment decisions: Through the role of the priority recommendation unit, the system can automatically recommend relatively better treatment plans, thereby optimizing treatment decisions. Doctors or patients can make quick decisions based on the recommended treatment plan, and further adjust the treatment plan by real-time monitoring of the patient's recovery progress. This function improves the efficiency and accuracy of the diagnosis and treatment process and ensures better treatment effects. Solve the complexity of treatment plan selection: When there are multiple treatment plans to choose from, the system can generate scientific treatment recommendations through multi-dimensional analysis. This effectively optimizes the decision-making process that relies on the doctor's experience in traditional treatment, and improves the scientificity and rationality of treatment.

[0101] Example 6

[0102] Please refer to Figure 1 ,Specifically: the error feedback module includes a collection unit and a time difference comparison unit;

[0103] The collection unit is used to recommend the drugs that are prioritized in the online recommendation module to the patient to be diagnosed, and monitor the recovery status of the patient to be diagnosed in real time when using the corresponding drugs to obtain the time when the patient to be diagnosed is fully recovered, and mark the time when the patient to be diagnosed is fully recovered as the actual time. .

[0104] The time difference comparison unit is used to determine the predicted time according to the dynamic conditional probability Dtgz of the earliest complete recovery extracted from the priority recommendation unit. , according to the predicted time , combined with the actual time obtained in the acquisition unit , determine the actual time difference Sc, the specific acquisition method is: ;

[0105] An error threshold W is set in advance, and the actual time difference Sc is compared and analyzed with the error threshold W to determine whether the treatment plan adopted by the current patient to be treated needs to be optimized. If the actual time difference Sc exceeds the error threshold W, it indicates that the current error is abnormal, and the optimization mechanism will be activated at this time; if the actual time difference Sc does not exceed the error threshold W, it indicates that the current error is not abnormal, and the optimization mechanism will not be activated at this time.

[0106] The optimization module is used to start and execute the optimization mechanism to adjust the original drug dosage taken in the treatment plan according to the actual time difference Sc. , in order to optimize the online treatment program, the specific adjustments are as follows:

[0107] ;

[0108] In the formula, represents the adjusted drug dose, Represents the learning rate, which controls the magnitude of each adjustment; Indicates the prediction time , Indicates the actual time difference, It represents the ratio of the actual time difference to the predicted time, which reflects the deviation between the predicted time and the actual time.

[0109] In this embodiment, in the acquisition unit, the system can accurately obtain the actual recovery time of the patient by real-time monitoring the recovery state of the patient to be diagnosed after using the recommended drug. This process provides important data support for subsequent error comparison and treatment optimization. By continuously collecting and updating the patient's recovery state, the system can timely evaluate the treatment effect and find out whether the treatment plan is effective, thereby providing the patient with a personalized optimized treatment plan. In the time difference comparison unit, the system calculates the actual time difference Sc by comparing the actual recovery time with the predicted time, and compares and analyzes it with the preset error threshold to determine whether the treatment plan needs to be optimized. This mechanism ensures the real-time and accuracy of the treatment, helps to discover potential treatment problems as early as possible, and avoids the lag of the patient's treatment effect. In the optimization module, the system adjusts the drug dosage in the treatment plan according to the actual time difference to optimize the treatment effect. This dynamic drug dosage adjustment mechanism can ensure that the patient gets a more appropriate drug dosage in time during the treatment process, thereby improving the treatment effect. Through this optimization mechanism, the patient's treatment plan will become more personalized and accurate, reducing the risk of overdose or underdose of drugs and improving the accuracy of recovery. In summary, the error feedback module ensures the dynamic adjustment of the treatment plan through real-time monitoring, time difference comparison, and optimization of drug dosage, and has the following main beneficial effects: Improved accuracy: By comparing the actual recovery time with the predicted recovery time, the treatment plan can be discovered and adjusted in time. Personalized optimization: Adjust the drug dosage according to the individual differences of the patient to optimize the treatment effect. Effect feedback mechanism: Through the feedback loop, the treatment plan is continuously improved to improve the treatment quality and the patient's recovery speed.

[0110] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An Internet online diagnosis and treatment data processing system based on a big data model, characterized by: It includes data collection module, association rule analysis module, online recommendation module, error feedback module and optimization module; The data collection module is used to collect relevant historical data of patients to be diagnosed and treated by using big data technology, and obtain the disease of the current patient to be diagnosed and treated through the online diagnosis and treatment process on the Internet. Based on big data technology, relevant historical data of patients with the same disease as the patient to be diagnosed and treated are collected, and a database is generated after data preprocessing; The association rule analysis module is used to build a big data model based on the database, analyze the association rules between the corresponding diseases and various types of drug treatments based on the big data model, so as to build a weighted lift degree Jtd, and determine the candidate pairs based on the weighted lift degree Jtd value; The association rule analysis module includes a sparse data processing unit and an association analysis unit; The sparse data processing unit is used to analyze the appearance status of each item set in the database according to the relevant historical data of non-target patients to calculate and obtain the weighted support Jzcd. The weighted support Jzcd is obtained by the following formula: ； In the formula, Reflects the frequency of co-occurrence of disease X and corresponding type of drug Y in the i-th item set, n represents the total number of item sets in the database, i=1, 2, 3, ..., n, represents the weight value of the i-th item set, represents the indicator function, which takes the value of 1 if the disease X and the corresponding type of drug Y in the i-th item set appear at the same time, otherwise it takes the value of 0; Among them, the weight is adjusted according to the sparsity of each item set in the database to obtain the weight value of the corresponding item set , where the weight value of the i-th item set is Take the example to calculate the weight value of the i-th item set. : ； The association analysis unit is used to analyze the association rules between the corresponding diseases and various types of drug treatments based on the relevant historical data of non-target patients and combined with the weighted support Jzcd to calculate the weighted lift Jtd. The weighted lift Jtd is specifically obtained by the following formula: ； In the formula, represents the frequency of disease X, represents the frequency of occurrence of the corresponding type of drug Y in the i-th item set; The online recommendation module is used to dynamically analyze the recovery status of various types of drugs in the candidate pairs in various time periods based on the candidate pairs and the relevant historical data of patients with the same disease as the patient to be diagnosed in the database, so as to construct the dynamic conditional probability Dtgz. Through the dynamic conditional probability Dtgz, the corresponding drugs are given priority recommendation. The error feedback module is used to recommend the drugs that are preferentially recommended in the online recommendation module to the patient to be diagnosed, and monitor the recovery status of the patient to be diagnosed in real time when using the corresponding drugs to obtain the time when the patient to be diagnosed fully recovers, and determine the actual time difference Sc according to the time when the patient to be diagnosed fully recovers; The optimization module is used to optimize the online treatment plan based on the actual time difference Sc.

2. An Internet online diagnosis and treatment data processing system based on a big data model according to claim 1, characterized in that: The data collection module includes a No. 1 collection unit, a No. 2 collection unit and a processing unit; The first collection unit is used to collect relevant historical data of patients to be diagnosed and treated by using big data technology, wherein the relevant historical data of patients to be diagnosed and treated include the medical records and treatment plans of patients to be diagnosed and treated in the historical period; and establish a communication relationship with patients to be diagnosed and treated through the online diagnosis and treatment process on the Internet, and combine the relevant historical data of patients to be diagnosed and treated to obtain the disease of the current patients to be diagnosed and the current relevant pathological data; The second acquisition unit is used to collect relevant historical data of patients with the same disease as the patient to be diagnosed based on big data technology, and to treat patients with the same disease as the patient to be diagnosed as non-target patients, where the relevant historical data of non-target patients includes the frequency of occurrence of disease X , the frequency of occurrence of the corresponding type of drug Y in the corresponding item set and the case where disease X and the corresponding type of drug Y in the corresponding item set appear at the same time; where the item set refers to the various types of drugs adopted for disease X; The processing unit is used to perform data preprocessing on the relevant historical data of the patients to be diagnosed and the relevant historical data of non-target patients to remove noise, identify and fill missing values ​​and abnormal values, and generate a database after dimensionless processing.

3. An Internet online diagnosis and treatment data processing system based on a big data model according to claim 2, characterized in that: According to the method of obtaining the weighted lift Jtd, the frequencies of the co-occurrence of the disease X and the corresponding type of drug Y in each set are calculated and marked as the weighted lift Jtd of each set. The association threshold K is pre-set and compared with the weighted lift Jtd of each set to determine the candidate pairs of patients to be diagnosed and treated, as follows: If the weighted lift Jtd of each item set exceeds the association threshold K, it indicates that there is an association rule between the corresponding disease in the current item set and the corresponding type of drug treatment. At this time, the corresponding disease in the current item set and the corresponding type of drug treatment are marked as candidate pairs; If the weighted lift Jtd of each item set does not exceed the association threshold K, it indicates that there is no association rule between the corresponding disease in the current item set and the corresponding type of drug treatment. At this time, the corresponding disease in the current item set and the corresponding type of drug treatment are not marked as candidate pairs.

4. An Internet online diagnosis and treatment data processing system based on a big data model according to claim 2, characterized in that: The online recommendation module includes a dynamic analysis unit, a data constraint processing unit and a priority recommendation unit; The dynamic analysis unit is used to extract the corresponding diseases and corresponding types of drug treatments involved in the candidate pairs from the relevant historical data of non-target patients according to the candidate pairs, and the relevant historical data of each period in the process of mutual matching and patient adoption, and combine the current relevant pathological data of the patient to be diagnosed and treated to dynamically analyze the recovery of each type of drug treatment in the candidate pairs in each period, and construct the dynamic conditional probability Dtgz. The dynamic conditional probability Dtgz is obtained by the following formula: ； In the formula, represents the joint probability of the non-target patient's recovery event, treatment plan, and the current severity of the patient's illness occurring simultaneously at time j; Indicates that at time j, non-target patients adopt the treatment plan And it has the probability of the severity of the current illness of the patient to be diagnosed; represents the recovery event of non-target patients at time j; represents the g-th treatment plan adopted by the non-target patient at time j; It indicates the current severity of the patient's condition at time j; j represents time; Among them, at time j, the recovery event of non-target patients Includes complete recovery, partial recovery, and no recovery; At time j, the gth treatment plan adopted by the non-target patient Refers to the original drug dosage taken when taking each type of drug for the corresponding disease and the time of taking; At time j, the current severity of the patient's condition The specific acquisition method is as follows: ； In the formula, Indicates the duration of the disease in the relevant pathological data, represents the frequency of disease attacks in the relevant pathological data, and Disease duration and frequency of disease onset The weight value of <1,0< <1, and The specific value is set by the user according to the situation.

5. An Internet online diagnosis and treatment data processing system based on a big data model according to claim 4, characterized in that: The data constraint processing unit is used to constrain the output value of the dynamic conditional probability Dtgz using the Softmax function to ensure that the recovery events of multiple groups of non-target patients are converted into probabilities, and the sum of the probabilities of the recovery events of the multiple groups of non-target patients is constrained to 1; The priority recommendation unit is used to obtain multiple groups of dynamic conditional probabilities Dtgz at different time points and different recovery events according to the method of obtaining the dynamic conditional probability Dtgz in the dynamic analysis unit, and extract the dynamic conditional probability Dtgz of the earliest complete recovery from them, so as to mark the candidate pair corresponding to the earliest complete recovery as a recommended pair, and determine the type of medicine in the recommended pair, so as to make priority recommendations to the patients to be diagnosed and treated.

6. An Internet online diagnosis and treatment data processing system based on a big data model according to claim 5, characterized in that: The error feedback module includes a collection unit and a time difference comparison unit; The collection unit is used to recommend the drugs that are prioritized in the online recommendation module to the patient to be diagnosed, and monitor the recovery status of the patient to be diagnosed in real time when using the corresponding drugs to obtain the time when the patient to be diagnosed is fully recovered, and mark the time when the patient to be diagnosed is fully recovered as the actual time. .

7. An Internet online diagnosis and treatment data processing system based on a big data model according to claim 6, characterized in that: The time difference comparison unit is used to determine the predicted time according to the dynamic conditional probability Dtgz of the earliest complete recovery extracted from the priority recommendation unit. , according to the predicted time , combined with the actual time obtained in the acquisition unit , determine the actual time difference Sc, the specific acquisition method is: ; An error threshold W is set in advance, and the actual time difference Sc is compared and analyzed with the error threshold W to determine whether the treatment plan adopted by the current patient to be treated needs to be optimized. If the actual time difference Sc exceeds the error threshold W, it indicates that the current error is abnormal, and the optimization mechanism will be activated at this time; if the actual time difference Sc does not exceed the error threshold W, it indicates that the current error is not abnormal, and the optimization mechanism will not be activated at this time.

8. An Internet online diagnosis and treatment data processing system based on a big data model according to claim 7, characterized in that: The optimization module is used to start and execute the optimization mechanism to adjust the original drug dosage taken in the treatment plan according to the actual time difference Sc. , in order to optimize the online treatment program, the specific adjustments are as follows: ； In the formula, represents the adjusted drug dose, Represents the learning rate, which controls the magnitude of each adjustment; Indicates the prediction time , Indicates the actual time difference, It represents the ratio of the actual time difference to the predicted time.

Images