Intelligent Neural Computing for Fractional‑Order Nonlinear Childhood Disease Modeling

This research presents the solutions of the fractional order childhood disease model by applying an intelligent computing neural network procedure. The model-based childhood disease is divided into three individual classes, susceptible, infected and recovered, which present a nonlinear model. The fractional order derivatives are used to provide more reliable results as compared to integer order. The construction of the neural network is presented via one hidden layer by taking fifteen neurons and activating sigmoid function. The training tests are performed through Bayesian regularization by selecting the data for training, validation, and testing, which is obtained through the Adams-Bashforth method. The neural network procedure’s correctness is approved via various tactics including results matching, small performances of absolute error, and best training performed as 10− 9, 10− 10, and 10− 12 for cases 1 to 3 of the model. It is also observed that fractional order results based on 0.9 are comparatively better in comparison with the 0.7 and 0.8, which shows that the fractional order derivatives perform well close to 1. Moreover, some tests using histogram errors, state transition, and regression have been used to authenticate the reliability of the scheme.