EVALUATION OF POLLUTION MORTALITY USING NARX AND NIO TIME SERIES PREDICTIVE ALGORITHM ON MATLAB

Mortality is an essential health effect of ambient air pollution and has been studied extensively. The earliest signal relates to fog occurrences, but with the advancement of more accurate methods of investigation and prediction, it is still possible to differentiate short-term chronological associations with day-to-day mortality at the historically low levels of air pollution now occurring in most developed countries. This paper studies and explores the methodologies for modeling, simulation, and controls in ANN-based on time series application of pollution mortality. To show and prove the effectiveness, simulated and operational data sets are employed to demonstrate the ability of neural networks in capturing complex nonlinear dynamics where NARX and NIO models are set up to explore and relate both steady-state and transient features on pollution mortality. The structures were configured, generated, and ran in MATLAB to create and train the platform. The validation, testing, and results validate that the techniques can be accurately applied, which implies both models effectively capture dynamics of the system up to a certain degree of acceptance. The associated parameters for the design and simulation are varied and set up according to the requirements which display that ANN can perform better than most conventional methods. Finally, it was established that NARX model outperforms more than the NIO model. Keywords— NARX, NIO, Regression, Time series prediction, Neural Networks, Machine Learning, Levenberg-Marquardt