Various methods and techniques have been resulted for forecasting of air pollution which is analytical, physical and numerical. The latest method for forecasting of air pollution is a model with a soft computing approach, that is neural network (NN) model. But in its application, the NN model used is still a standard NN model that has not been combined with wavelet methods or fuzzy techniques. This research is based on the characteristics of meteorological and pollution data including time series data and the superiority of the NN and wavelet methods in part or in combinations. This research developed  a feed forward wavelet neural network (FF-WNN) model based on  the superiority of the B-spline and Morlet wavelets as an activation function in the FFNN model to forecast air pollution problems based on meteorological data variables and the air pollution parameters. The application of this model in modeling of air pollution in the city of Mataram-Lombok, West Nusa Tenggara-Indonesia gave significant results based on the root mean square error (RMSE) indicator. Considering that Lombok Island is a small island, this model can be representation of air pollution modeling for a city with characteristics as a city on a small island.

wavelet neural network, pollutant, air pollution, forecasting of time series, small island.