Research On Recognition Algorithm Of Alzheimer’s Disease Based On The Transfer Of Few Feature Samples

Alzheimer’s disease is the first major neurodegenerative disease.This article focuses on the use of public data sets with a large number of samples to solve the contradiction between inadequate Alzheimer’s samples in a small number of specific areas and the excellent classification performance of the model through transfer learning.Therefore,the research content of this article is as follows:First,it summarizes and analyzes the relevant theories and technologies in casebased transfer learning.Based on the existing theories and technologies,combined with the particularity of the data used in this article,that is,the characteristics of the open data set(source domain)are few and the local data set(Target domain)has many features.We consulted the information about the technology related to feature growth and laid a solid theoretical foundation for the subsequent processing of this data.Secondly,due to the characteristics of few source domain features and many target domain features,the general transfer learning method cannot be applied to the problem in this scenario.A feature growth algorithm for K-nearest neighbor adaptation using common features is proposed to solve the feature mismatch problem between the source and target domains.The feature growth algorithm of K-nearest neighbor adaptation is based on the distance relationship between the common features of the two data set samples,and selects suitable features from the target domain samples and fills them into the source domain,so that the source domain and the target match the sample features.The simulation accuracy of the algorithm in different dimensions is analyzed through experimental simulation,and the effectiveness of the proposed algorithm is verified.Finally,to further reduce the gap between the source and target domains after feature matching,a cross-domain metric manifold alignment algorithm is proposed.This algorithm fully considers the manifold alignment between samples in the data transformation process.By adding label information,the manifold structures between similar samples between domains and domains are close to each other,and the manifold structures between different samples are far away from each other.Experimental results show that compared with traditional algorithms,the proposed algorithm can effectively diagnose Alzheimer’s disease patients in the scene where the source and target domains are distributed differently.Based on the above research,this paper implements a classification algorithm from data preprocessing to early diagnosis of Alzheimer’s disease,and verifies the effectiveness of the proposed algorithm through experiments.