Research On Method Of Tumor Feature Gene Section With Fuzzy Neighbourhood

In biomedicine,DNA microarray technology can be used to obtain a large number of gene expression profile data.Researchers construct effective tumor classification models by analyzing genetic data,and these models have important research significance and application value for clinical diagnosis and treatment of tumors.The gene expression profile data can approximate the expression information of the whole genome of the biological tissue cells.However,the data is characterized by high dimensionality and small samples,and contains a lot of redundancy information,which will affect the diagnosis if not processed.Therefore,it is the core problem of gene expression profile analysis to study the accurate description of feature genes and its correlation with tumor classification and select effective feature genes from thousands of genes,and is also the research focus of this paper.In the process of feature gene selection,many existing algorithms pay more attention to the selection and research of feature genes measurement methods,while ignoring the accurate description of the original information in data processing.By analyzing the gene expression profile data,this study introduces the neighborhood relationship and fuzzy relationship of rough set theory into conditional entropy in different ways to construct the corresponding feature genes selection model.Because of the combination of the algebraic definition and information theory definition of the significance of feature genes subset,the measurement mechanism has become more perfect.At the same time,the performance of traditional gene selection methods is improved,and a superior feature genes subset is obtained.The main innovations are listed as follows:(1)In the genetic data,there are many data that the sample neighborhood is not completely included in its decision equivalence class.For the problem that such key data description is not accurate enough,this paper constructs a fuzzy neighborhood conditional entropy model.Firstly,fuzzy neighborhood granule and fuzzy decision making are used to characterize gene expression profile data more accurately and reduce the loss of original information during computation.Then,the definition of fuzzy neighborhood conditional entropy and the proof of its theorems,such as monotonicity,are given based on the proposed model.To tolerate noise of the data,the parameters are set and its selection is discussed.Finally,the importance of candidate feature genes is evaluated by the monotonicity principle of the proposed model to obtain an effective feature genes subset.The experimental results show that the proposed method can eliminate redundant and noisy data effectively and achieve better tumor classification results.(2)In order to overcome the problem of sample misclassification in the process of feature genes selection,this paper proposes a rough uncertainty measure model based on work(1).Firstly,the combination of fuzzy similarity relation and neighborhood radius of rough set theory is used to construct the fuzzy neighborhood granule of the sample,and the rough decision is defined by the fuzzy similarity relation between the samples and the decision equivalence class.Then,the fuzzy neighborhood granule of the sample and rough decision are introduced into the conditional entropy in a more targeted way,and a rough uncertainty measure model is proposed.Since the noise and redundancy in the data are allowed to exist,this paper introduces the variable precision model and analyzes the selection of parameters.Finally,based on the proposed model,a feature gene selection algorithm is designed to select effective genes.The experimental results show that the method can achieve higher classification accuracy with fewer feature genes.