Research On Association Classification Algorithm And Its Application In Diagnosis Of Coronary Heart Disease

With the development of social economy and the continuous abundance of material conditions,the national lifestyle has become increasingly unhealthy.In the context of the accelerating aging of society and urbanization,cardiovascular diseases represented by coronary heart disease have become popular in China in recent years,and have become the leading cause of death among residents.In addition,the continuous development of information technology and physical storage technology has made the methods of medical diagnosis more diverse and the process more complicated,thus accumulating a large amount of medical data.How to effectively use these data and how to find valuable information to provide reference for disease prevention and diagnosis have important research significance.The specific research results of this paper are as follows:Aiming at the shortcomings of existing association classification algorithms,such as large resource consumption,difficult rule pruning and complex classification model,an improved scheme ACCP based on partition mining and prior pruning is proposed.The former items of the classification rules are mined according to the different classification attribute values,and the frequent item set mining process and rule pruning process are improved and optimized.The experimental results based on UCI dataset show that the improved scheme has better classification performance than the traditional CBA association classification algorithm and C4.5 decision tree algorithm.The average classification accuracy is increased by 3.93 and 5.4 percentage points respectively and the average sensitivity is increased by 3.95 and 4.51 percentage points respectively,and the running time is obviously better than the traditional CBA algorithm.The improved scheme has achieved good application results.Aiming at the problem that the traditional Relief series algorithm can not filter redundant features,the FSRMI feature selection algorithm based on Relief F algorithm and mutual information is proposed.On the basis of the original Relief F algorithm,the FSRMI algorithm abandon the method of eliminating the invalid features by setting the feature weight threshold,and adopt the heuristic feature reduction method based on mutual information.The feature subseis is sequence forward searched by calculating the mutual information of the feature subset and the category attribute,and the consequence of whether the calculated result reaches the mutual information of the feature complete set and the category attribute is used as the termination condition of the feature subset generation process.Finally,a quadratic reduction based on mutual information feature selection is performed on the generated feature subsets.The redundant features of the feature subset are removed by calculating the information metric MIFS,and a better dimensionality reduction effect is achieved.The experimental results show that the attribute reduction rate of the model on the UCI data set Breast reaches 44.4%,and the classification performance is further improved compared with the attribute complete set.Based on the actual collected coronary heart disease dataset,the performance of the proposed ACCP association classification algorithm and FSRMI feature selection algorithm in the diagnosis of coronary heart disease were verified.Firstly,the data set is preprocessed by data filtering,missing value padding and data discretization.Then,based on the FSRMI feature selection algorithm,a feature subset containing 11 features is obtained.Finally,a series of comparative experiments were carried out on the processed coronary heart disease dataset,which proved the practicability and effectiveness of ACCP association classification algorithm in the diagnosis of coronary heart disease.In addition,based upon the ACCP classification algorithm,the verification experiments are carried out on the dataset before and after the FSRMI feature selection,which proves that the FSRMI feature selection algorithm can effectively reduce the redundancy of the feature set and has a good application effect.