A Mild Depression Recognition with Classifier Combination Method Based on Differential Evolution

Depression is one of the most common mental disorders affecting people, but its recognition rate is low due to subjectivity and other factors. Mild depression, in particular, has milder and less recognizable symptoms. In this study, we used electroencephalography (EEG) and machine learning methods to identify mild depression. Our approach was to construct a new weighted combinatorial classifier model to distinguish patients with mild depression from normal controls. In this experiment, 10 mildly depressed patients and 10 normal controls watched different emotional facial pictures, recorded their EEG signals and preprocessed them, and then extracted linear and nonlinear features to construct feature vectors. K-nearest neighbor(KNN), support vector machine(SVM), logistic regression(LR), random forest(RF) and back propagation neural network(BPNN) were selected as five individual classifiers, and the differential evolution algorithm(DE) was used to optimize the weights to improve the overall performance of the recognition model. The experimental results showed that the classification accuracy of the proposed method was better than that of the individual classifiers, and the highest 99.09% was achieved when the number of iterations was 50, indicating that the fusion model had a higher recognition accuracy of mild depression than the single modes. At the same time, compared with other combination strategies, this method was also better than other strategies. This research may provide a means to identify mild depression.