PRUNED LAZY LEARNING MODEL FOR THE TIME SERIES PREDICTION: APPLICATION TO PREDICTION OF THE GREAT SALT LAKE

The relationships between hydrologic variables are often nonlinear. Usually the functional form of such a relationship is not known as a priori. There are many techniques. They are linear methods which include AR and ARIMA, and nonlinear methods such as artificial neural networks. In general, these methods try to build up a model of the process. The model is then used on the last values of the series to predict future values. The common difficulty to all the methods is the determination of sufficient and necessary information for a good prediction. A learning methodology is provided here for approximating the underlying the nonlinear function. The nonlinear time series learning model which is introduced in this study is based on a linear piecewise approximation method called the Lazy Learning (Aha, 1997; Bontempi et al., 1999) and it is applied to the Great Salt Volume time series. This method dose not suffer from the problem of the local minima (in example, some neural networks give the minimum which is not global, however a local one instead).