Research On Risk Assessment Model Of Cardiovascular Diseases Based On Data Mining

Chinese cardiovascular disease report 2015 reported that cardiovascular disease(CVD)is still the first cause of death among Chinese residents in 2014.Due to the high cost of cardiovascular disease and irreversible damage to human bodies,people are gradually shifting the focus from the treatment to the prevention.Therefore,it is urgent to put forward a simple and effective method to evaluate the risk of cardiovascular disease by using the health examination data.This model includes not only the independent risks for cardiovascular disease but also include the geographical features,people's habits,the latest research progress and any other factors.It has strong pertinence and keeps pace with the times.With the standardization and popularization of health examination,the medical industry data explodes,causing the large cross-section data.Traditional research methods based on longitudinal time series data are limited because of the lack of long data accumulation.With the rise of machine learning,some methods of Data mining can be used to solve this problem.Based on the analysis of medical industry data and specific data mining algorithms and combined with medical expert experience,this thesis established a cardiovascular risk assessment model and the methods as well as procedures of risk assessment of cardiovascular disease were given.The model ensures that the input of the model are all structured variables by using a simple interpretation of the ECG signal and its transformation structured input as well as other subjects processing results.The input of model was determined with the risk of cardiovascular disease in the clinical concern,association rules analysis of relevant indicators and conditions using Apriori algorithm.The cardiovascular risk assessment model was constructed by Logistic regression analysis,judged by the accuracy rate,based on the idea of system identification in this thesis,the construction of the evaluation model is studied and the following results are obtained:Firstly,an improveed wavelet packet denoising algorithm is proposed,according to the characteristics and distribution of ECG signal,waveform processing requirements and signal to noise ratio SNR.This algorithm has the advantages of wavelet packet and threshold algorithm comparing with several other common denoising methods.Secondly,the ECG signal is converted into structural variables.The anti-jamming performance of the processing algorithm and the characteristic detection rate improves by using wavelet modulus maxima algorithm to extract the singular points of the signal.Feature points and feature intervals of ECG signals are extracted combined with signal characteristics.The interpretation results of ECG signal as one of the model input comes from the comparison between extraction of features and Normal ECG signals combined with the experience of medical interpretation rules.Finally,the risk assessment model of cardiovascular disease is built based on objective facts.The improved interestingness Apriori algorithm is used to mining the train set in association rules.It can help the study find the relationship between indicators and disease accurately.It can also help to determine input of the model and build the basic model.The method of improved Logistic regression analysis is used to build the model which can assess the risk of cardiovascular disease.The research results of this thesis will be applied in the comprehensive health service platform of the subject.The research of this thesis applied in the subject of Integrated Health Service Platform.