Performance of Levenberg-Marquardt neural network algorithm in air quality forecasting

Abstract

Levenberg-Marquardt algorithm and conjugate gradient method are frequently used for optimization in multi-layer perceptron (MLP). However, both algorithms have mixed conclusions in optimizing MLP in time series forecasting. This study uses autoregressive integrated moving average (ARIMA) and MLP with both Levenberg-Marquardt algorithm and conjugate gradient method. These methods were used to predict the Air Pollutant Index (API) in Malaysia’s central region where represent urban and residential areas. The performances were discussed and compared using the mean square error (MSE) and mean absolute percentage error (MAPE). The result shows that MLP models have outperformed ARIMA models where MLP with Levenberg-Marquardt algorithm outperformed the conjugate gradient method.