Health Big Data Classification Using Improved Radial Basis Function Neural Network and Nearest Neighbor Propagation Algorithm

• Published in: IEEE access Vol. 7; pp. 176782 - 176789
• Main Authors: Jiang, Congshi, Li, Yihong
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 2019; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Accuracy; Algorithms; Big Data; Cardiovascular disease; Classification; Classification algorithms; Clustering; Clustering algorithms; Convergence; convergence time; Datasets; Diabetes mellitus; Diagnosis; Diseases; Exponential functions; Health big data classification; Heart diseases; manifold analysis; Manifolds; Medical diagnostic imaging; Medical personnel; neighbor propagation algorithm; Neural networks; Optimization; Physicians; Propagation; Radial basis function; radial basis function neural network; similarity matrix; Subtraction; Tuberculosis
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2019.2956751

Health big data classification can effectively improve the level of medical and health services and management, help medical staff to carry out auxiliary diagnosis, improve the efficiency of doctors and the accuracy of diagnosis. To solve the problem of health big data classification, this paper proposes a Radial Basis Function (RBF) neural network classification algorithm based on manifold analysis and nearest neighbor propagation (AP) algorithm. First, the data set is processed by manifold analysis algorithm. Then, the similarity matrix is adjusted by exponential function, and AP clustering is carried out again. On this basis, RBF neural network classifier is constructed. In order to improve the classification accuracy and shorten the convergence time, an algorithm for constructing the variable basis width neural network model is proposed. This method is based on the subtraction clustering algorithm and K-means algorithm to determine the cluster center. The maximum distance between the sample and the cluster center is selected as the base width. The base width is updated adaptively with the optimization of the cluster center. Finally, three data sets of patients with coronary heart disease, diabetes mellitus and bronchial tuberculosis were collected as test data, and the accuracy of classification and convergence time were compared. Experimental results show that this method can improve the classification accuracy and convergence speed of large data sample set.