Summary of the invention
It is an object of that present invention to provide a kind of big data digging systems for tcm clinical case information, in order to overcomeThe method of traditional manual sorting data is no longer met the Research Requirements of medical research personnel and preferably saves, believed using medical recordThe problem of breath, is designed by using C#.net environment and B/S architecture technology, and the research for meeting medical research personnel needsIt asks, and preferably saves, using medical record information, established for the further new rule of research tcm clinical data mining, new methodBasis is determined.
In order to solve the above technical problems, the present invention adopts the following technical scheme that: one kind is believed for tcm clinical caseThe big data digging system of breath, the system include: medical record management module, data source modules and data-mining module；Wherein, instituteStating medical record management module allows doctor's typing medical record and retrieves to medical record information, and provides data to the data source mouldBlock；The data source modules are responsible for storing the data that the medical record management module is submitted and other external datas are gone forward side by side professional etiquette modelChange processing；The data-mining module is responsible for the data transmitted for the data source modules and carries out data prediction, knowledge diggingPick and knowledge analysis processing.
Further, the medical record management module is made of medical record typing submodule and medical record retrieval submodule；Wherein,Medical record typing submodule includes essential information typing and case history input function；Medical record retrieval submodule is mainly responsible for inquiry and showsThe medical record information of patient, and provide and accurately inquired by patient's identification number, by functions such as name of disease inquiry, feeling attending doctor inquiries.
Further, the mining process of the data-mining module point: data prediction, Extracting Knowledge and knowledge analysisThree phases.
The present invention have compared with prior art it is below the utility model has the advantages that
The present invention program no longer meets the Research Requirements of medical research personnel for the method for traditional manual sorting dataAnd preferably save, using medical record information the problem of, be designed by using C#.net environment and B/S architecture technology, it is fullThe foot Research Requirements of medical research personnel, and preferably save, using medical record information, for further research tcm clinicalThe new rule of data mining, new method are laid a good foundation.
Specific embodiment
With reference to the accompanying drawing and specific embodiment to the present invention carry out in further detail with complete explanation.It is understood thatIt is that described herein the specific embodiments are only for explaining the present invention, rather than limitation of the invention.
Referring to Fig.1, a kind of big data digging system for tcm clinical case information of the present invention, the system include:The system includes: medical record management module, data source modules and data-mining module；Wherein, the medical record management module allowsDoctor's typing medical record and medical record information is retrieved, and provides data to the data source modules；The data source modules are negativeDuty stores the data that the medical record management module is submitted and other external datas and carries out standardization processing；The data miningModule is responsible for the data transmitted for the data source modules and carries out data prediction, knowledge excavation and knowledge analysis processing.
Referring to Fig. 2, medical record management module is designed based on reality needs, and medical record management module includes two functions: diseaseCase typing and medical record are retrieved.Medical record typing is divided into essential information typing and case history typing again, and essential information specifically includes that medical recordNumber, name, gender, occupation, name race, wedding condition, the age, morbidity solar term, identity card, phone, contact address, past medical history, family historyThis 13 attributes；Case history typing is mainly the medical record information of typing patient, is mainly had: patient's identification number, main suit, present illness history, diagnosis,Therapy, Chinese medicine four methods of diagnosis information, laboratory check information etc..The function also comprising printout, medical record are retrieved simultaneously for medical record retrievalThe medical record information for mainly inquiring and showing patient provides and accurately inquires by patient's identification number, inquires by name of disease inquiry, feeling attending doctorDeng.
For medical record typing, a patient's identification number is created, typing essential information, case information, defeated based on what is designed respectivelyEnter information model, the option that system display needs to input, when inputting medical record data, to carry out unified standard, system can be rightSome data provide prompt and default value.Such as, the input of Time of Day.Doctor can all data of gradation typing, and with inputPatient's identification number facilitates data input and updates as mark.Additionally set up data dictionary table, the knowledge of some normalizations of typing, such asChinese medicine name and number, symptom and coding, card type and coding after specification etc., are deposited into database.
Medical record is retrieved, system provides " precise search " and " fuzzy query " and retrieves two ways, by that can setDifferent condition, such as name, age, patient's identification number etc., input inquiry condition, doctor can retrieve all medical records of needsInformation, default operating right, doctor do not have modification authority to other people diagnosis and treatment medical records.After providing medical record search function, doctorLife can check patient's medical record in advance, can propose more reasonable diagnosis and treatment opinion when reserving further consultation so as to patient, can also be to oneselfThe place neglected before is corrected, and young doctor can also enrich the diagnosis of oneself by checking the medical record of experience doctor's typingKnowledge.System supports printout.
Referring to Fig. 3, the data-mining module is designed based on Research Requirements, is intended to carry out knowledge to medical record dataIt excavates, mining process is divided into following three phases:
Data preprocessing phase is imported by database obtain data first, then through number in process of data preprocessingData preprocess carries out arrangement specification, quantization modulation etc. to medical record information, obtains data available.Main flow includes: by " dataCleaning " carries out the operation such as deletion default value, deletion error to the data of selection；By " data conversion ", SQL statement is calledOperable data in machine learning are converted by the data dictionary table in the data control knowledge base of medicine expression.Simultaneity factorData transverse and longitudinal transformation function is provided, by the data deposit data mining library after conversion after having handled, is made with being studied to the later periodWith.Mode there are also a kind of pair of data prediction is exactly to pass through Algorithm for Attribute Reduction to carry out reduction to data attribute: being adjustedThe attribute information after reduction is extracted with MIBARK algorithm, is stored in data mining library.
In the Extracting Knowledge stage, the data mining duty of realization has: association rule mining, neural network classification prediction, Yi JiqiHe selects to need Extracting Knowledge type, after specification handles data mining capability in data mining knowledge process, symptom data collection, card type data set, Chinese medicine data set, difference maintenance data mining algorithm, to association analysis knowledge and nerveNetwork class prediction knowledge is excavated.In association analysis knowledge excavation, the algorithm file of calling system encapsulation passes through settingSupport threshold, confidence threshold value parameter obtain the correlation rules of needs.System is by frequent item set and each frequent item setStatistical counting, be shown on the page, association results are with the sequence list display of " former piece=> consequent support confidence level ".?During neural network classification prediction knowledge excavates, according to the parameter set: error, learning rate, maximum number of iterations, each node layerNumber, training sample number etc., establish out prediction model of classifying.Then the sample of selection test is predicted, after system operation,Export discrimination, recognition result.
In the last knowledge analysis stage, by the Extracting Knowledge combination knowledge in knowledge base of acquisition, medicine side is excavatedIncidence relation and symptom-card type dialectical discrimination rule between agent Compatibility Law, symptom and medication.
The algorithm being designed into for the big data digging system of tcm clinical case information has:
MIBARK algorithm based on Importance of Attributes is a kind of by introducing between the condition class and Decision Classes of decision tableMutual information measures the old attribute reduction algorithms of Importance of Attributes.MIBARK algorithm description is as follows: input: decision table, wherein U is indicatedTransaction data set (TDS), C indicate that conditional attribute, D indicate decision attribute.Output: property set B is conditional attribute collection C relative to decision categoryA Relative Reduced Concept of property collection D.In data preprocessing phase, system extracts asthma primary symptom shape using the algorithm, eliminates redundancySymptom is prepared for Medical Records data mining in next step.
The computer of fast reaction based on to(for) bit string, improves Apriori algorithm: will be in the D of item data libraryEach things I is indicated with a bit string.Occur being 1, does not occur being 0.Algorithm only needs run-down database after improvement, generatesInitial item bit string operates the logical "and" of item bit string, a support meter is determined by the number of " 1 " in statistical items bit stringNumber.
Also use a kind of improved BP neural network algorithm in this system, the innovatory algorithm by the competition learning of hidden layer withThe adaptive adjustment of learning rate avoids falling into local minimum to make algorithm fast convergence.The basic thought of algorithm: hidden layer has been calculatedAfter the error of each node, the weight for the node for having worst error is normally corrected, and to the weight of other units all to phaseOpposite direction amendment；After each algorithm iteration is complete, calculates the value of error function and be compared with previous value, if errorThe value of function increases, then represents toning learning rate, learning rate should be lowered in next iteration with certain ratio, if error letterSeveral values reduces, then representing learning rate amplification can increase.System is in asthma symptoms-matched knowledge excavation of card type, firstUsing MIBARK algorithm, main sympotomatic set is extracted, then uses improved back-propagation again, classification prediction model is established, to testData carry out classification prediction, and then excavate the matching grating between asthma symptoms and card type
The above is not intended to restrict the invention, and to those skilled in the art, the present invention can have various changeDynamic and variation.All any modification, equivalent replacement, improvement and so within the spirit and principles of the present invention, should be included inWithin protection scope of the present invention.