Autism Spectrum Disorder (ASD) has a significant impact on the health of patients, and early diagnosis and treatment are essential to improve their quality of life. Machine learning methods, including multi-classifier fusion, have been widely used for disease diagnosis and prediction with remarkable results. However, current multi-classifier fusion methods lack the ability to measure the belief level of different samples and effectively fuse them jointly.
To address these issues, a multi-classifier fusion classification framework based on belief-value for ASD diagnosis is proposed in this paper. The belief-value measures the belief level of different samples based on distance information (the output distance of the classifier) and local density information (the weight of the nearest neighbor samples on the test samples), which is more representative than using a single type of information. Then, the complementary relationships between belief-values are captured via a multilayer perceptron (MLP) network for effective fusion of belief-values.
The experimental results demonstrate that the proposed classification framework achieves better performance than a single classifier and confirm that the fusion method used can effectively fuse complementary relationships to achieve accurate diagnosis.
Furthermore, the effectiveness of our method has only been validated in the diagnosis of ASD. For future work, we plan to extend this method to the diagnosis of other neuropsychiatric disorders.