COMPARISON OF TWO FORECASTING METHODS IN TIME SERIES DATA WITH SEASONALITY

This paper describes two forecasting methods in time series data with seasonality. The first method is an exponential smoothing model (parametric model) and the second forecast method is a machine learning model (artificial neural network model). We used a time series data with seasonality such as sunspot number data to evaluate the models. Our experiments show that the second forecast method has a better result in the sunspot data. We have also understood the difficulty in the modeling and implementation of those methods to forecasting and discuss their use in a real world application. Correlation of low season of sunspots and the low market prices is also observed.