Research Of Cardiovascular Health Status Prediction Model Based On Data Mining

According to statistics, chronic diseases has become the first cause of death in our population,as the dangers of Cardiovascular disease continues to grow in the global world, people’s aspirations for Cardiovascular health continues to raise, if we can effectively evaluate and predict the Cardiovascular health, it will play a great role in Cardiovascular disease prevention. Hospital physical examination center organizes physical examination without involving life features data related on cardiovascular health, so traditional forecasting methods based on time series and conventional medical standards are not applicable. If we predict Cardiovascular health status level by the queue data, data mining can make up for the deficiencies of traditional data analysis methods, and fully explore the laws and predict the Cardiovascular health level.Then this paper describes the prediction methods and processes of Cardiovascular health status based on the data mining process. We focus on how to get physical characteristics by attribute reduction and how to get cardiovascular health status sequence, physical characteristics and health status sequence combine to explore to construct a model for the prediction, and finally we get the prediction model of Cardiovascular health status rating based on RBF neural network-based. Therefore, this paper focus on the course to predict Cardiovascular health status data, problems need to solve and goals to achieve to do the research work, the main research work of this paper can be summarized as follows:1. A model constructed to discriminate and predict health status of the data mining process. Based a specific requirements analysis with the already acquired examination queue data and the target, We need to do data extraction and conversion and build a data mining model to predict Cardiovascular disease or health status.2. A new algorithm to fuse filter and wrapper is proposed for selecting features. The attribute reduction method being introduced involves feature conversion and feature selection. Different selection algorithms based on filter and wrapper are talked in the paper and compare their advantages and disadvantages, in order to reduce redundancy and to improve performance, a new algorithm to fuse filter and wrapper is proposed for selecting features.3. Based on the fuzzy clustering to distinguish health level. Screened characteristic feature selection based on fuzzy logic expounded inadequate reasoning and hard clustering algorithms determine the health status and the final choice of fuzzy logic clustering algorithm to achieve a healthy state the reasons for grading; Then this paper presents the disadvantages based on fuzzy logic reasoning and hard clustering algorithms to determine the health status, and the reason of the final choice of fuzzy logic clustering algorithm to achieve to evaluate healthy state for grading based on features selected in the last process.4. A predictive model based on RBF neural network. A predictive model is constructed using the static physical health status indicators and dynamic health status resulting sequence discriminated by evaluation model introduced by last module based on RBF neural network.Real data experimental simulation were carried out on the three modules for the above process using six consecutive years’ adult health examination data as experimental data for training and testing from two hospitals By analyzing the experimental results, this model has a certain effectiveness and practicality. In the process of the experiment verification, we first analyze the already acquired examination queue data structure and existing data issues. In view of the existing data problem, we use related cleaning methods aim at the different causes of abnormal data. All the cleaning work lay a good foundation for the next step of the data mining process. And this paper introduces and analyses source code based on different algorithms module above data mining process, and finally to realize a model for predicting cardiovascular health status level by integrating different modules Weka source functions and interface.