粗糙集属性约简方法在医疗诊断中的应用研究

发布时间：2018-01-05 10:22

本文关键词：粗糙集属性约简方法在医疗诊断中的应用研究　出处：《苏州大学》2015年硕士论文　论文类型：学位论文

【摘要】：如今,医院积存了海量的医疗诊断数据,如何利用先进的数据处理技术对其开发利用,辅助医生诊断,已经成为当今医疗事业发展的一个重要方向。目前医疗诊断的主要方法是根据患者的症状进行临床诊断。但随着疾病种类的增多,症状之间的干扰性大大增强,这给医生带来很大负担。本文主要针对医疗诊断问题及粗糙集属性约简算法进行研究,提出了两种属性约简算法辅助医疗诊断。主要工作如下:1)针对HORAFA算法以及现有改进算法所获得的结果中经常出现属性权值相同的问题,本文结合医疗诊断数据特征,改进算法的启发规则及属性删除操作,提出一种改进的基于差别矩阵的启发式约简算法。实验结果表明,改进算法能够提高约简效率,获得更优约简。2)为了解决大规模医疗诊断数据的约简问题,提出一种自适应遗传约简算法。该算法利用改进的属性权值构造个体适应度函数;使用新的最优选择策略丰富群体种类,避免陷入局部极值;引入属性相似度概念减少交叉操作,且降低了适应度函数的计算次数;改进变异操作,避免个体中存在权值相同的属性。仿真实验表明,该算法在属性数目繁多、数据量庞大的信息系统中更适用。3)基于上述的遗传约简算法,对2型糖尿病数据进行属性约简,提取关键症状,辅助医生诊断,减少误诊和漏诊。
[Abstract]:Now, the hospital has accumulated a large amount of medical diagnosis data, how to use the advanced data processing technology to develop and use to assist doctors in diagnosis. At present, the main method of medical diagnosis is to make clinical diagnosis according to the symptoms of the patients. But with the increase of the types of diseases, the interference between the symptoms is greatly enhanced. This brings a great burden to doctors. This paper mainly focuses on medical diagnosis problem and rough set attribute reduction algorithm. Two attribute reduction algorithms are proposed to assist medical diagnosis. The main work is as follows: 1) aiming at the problem of the same attribute weight value in the results of HORAFA algorithm and existing improved algorithm. In this paper, an improved heuristic reduction algorithm based on discriminant matrix is proposed by combining the features of medical diagnosis data, the heuristic rules and attribute deletion operations of the algorithm are improved. The improved algorithm can improve the efficiency of reduction and obtain better reduction. 2) in order to solve the problem of large-scale medical diagnosis data reduction. An adaptive genetic reduction algorithm is proposed, in which the individual fitness function is constructed by using the improved attribute weights. The new optimal selection strategy is used to enrich population types and avoid falling into local extremum. The concept of attribute similarity is introduced to reduce crossover operations and to reduce the number of computation of fitness function. The simulation results show that the algorithm is more suitable in the information system with a large number of attributes and large amount of data. 3) based on the above genetic reduction algorithm. The data of type 2 diabetes were reduced, the key symptoms were extracted, the diagnosis was assisted, and the misdiagnosis and missed diagnosis were reduced.
【学位授予单位】：苏州大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：R4;TP18

【参考文献】