Research On Depression Recognition Based On Case-based Reasoning

In recent years,the incidence of depression has increased significantly in different countries,and the resulting damage has also increased year by year.At present,the main methods for depression recognition are: Doctor's subjective recognition method,which is easy to deviate from the objective fact;Body hormone assay method,which is often confused with some of the physical diseases and causing misdiagnosis,and it takes longer time and less convenience for patients to wait for results.In order to make the depression recognition process more objective,reliable,efficient and convenient.Based on EEG signals,this thesis studied the method of depression recognition by using case-based reasoning.The main contents are as following:(1)To improve the recognition accuracy and generalization ability of the weighted similarity search strategy in case-based reasoning model,a novel fusion determining weights algorithm based on genetic algorithm is proposed,which is combined principal component analysis method,variation coefficient method and information entropy method.The search strategy utilized the weighted standard Euclidean distance and improved cosine similarity to calculate the similarity between samples.The novel search strategy based on fusion determining weights and its similarity calculation method is verified by using 5-fold cross validation method on twenty-three public datasets provided by UCI.The results demonstrated that the proposed method can effectively improve the accuracy of recognition under different datasets.(2)The proposed search strategy based on fusion determining weights and its similarity calculation method was applied in depression recognition.First,information retrieval table was built by reducing attribute of EEG data.Then,the weights of attribute in the table was determined.At last,the search strategy was determined by comparing the different similarity calculation methods for depression recognition under the fusion determining weights algorithm.The results showed that the proposed search strategy based on fusion determining weights algorithm and its standard Euclidean distance similarity calculating method can obtain approximately 91.64% accuracy in depression recognition and can effectively reduce the phenomenon of misdiagnosis and missed diagnosis to a certain extent.This thesis has improved the recognition accuracy and generalization ability of the search strategy in case-based reasoning model.It provided a new idea for the depression recognition and provided a certain reference for the relevant areas of depression research.