Fuzzy Rough Set Based Decision Methods And Their Applications In Medical Field

Medical decision-making based on large-scale complex data is a difficult problem in artificial intelligence and medical field.Traditional decision-making methods are for the small-scale data,however for the large-scale sample and high dimension data which are ineffective.Meanwhile,it is difficult for existing machine learning classification algorithms to make decisions efficiently and accurately.The decision-making method based on fuzzy rough set combines the advantages of rough set and fuzzy set,which can deal with fuzzy uncertain and incomplete information well,and is widely used in intelligent decision-making.Therefore,based on the fuzzy rough set theory,this paper first studies the feature selection method for medical large-scale data,and then extracts the decision rules from the data after attribute compression.Based on the fuzzy rough set theory,a method for attribute reduction of fuzzy discernibility matrix is proposed for genetic data with high attribute dimension and few sample objects,and a heuristic algorithm(FDM)is designed.The fuzzy discernibility matrix is a fuzzy generalization of the discernibility matrix,which can show the difference in the degree of discrimination between objects with the same attributes,and select the attributes with higher degrees of discrimination in priority,so as to improve the performance of the classification learning algorithm.On the colon cancer data,FDM algorithm effectively screened out 5 key genes related to the incidence of colon cancer from 2000 genes,the classification accuracy increased from 74.17%to 88.06%.According to large sample size of the clinical diagnosis data,the feature selection method of fuzzy related family based on fuzzy covering rough set is proposed in this paper,and the corresponding heuristic algorithm(FRF)is designed.In the numerical experiment,compared with the existing three representative attribute reduction algorithms,the maximum average running time on 7 public data sets was compressed from 2858.77 seconds to 29.21 seconds over FRF algorithm,and the classification accuracy was maintained.In terms of thyroid data,the relevant indicators of thyroid disease diagnosis were screened in 45.9 seconds,and the classification accuracy was improved from 93.37%to 96.69%The fuzzy discernibility matrix and fuzzy related family attribute reduction method proposed in this paper efficiently extract the key diagnostic indicators,improve the data quality,and improve the accuracy of intelligent diagnosis,which is expected to assist doctors in clinical diagnosis.