Week 8

Resample Models

Motivation to create models with resample data is to find more than one time step ahead, without excluding last known water level data which has highest correlation.

Methodology is to resample data keeping last known data set. So , accordingly if you need to predict two hours ahead, data set is resample with sample time of 2 hours keeping last known data fixed and modelling the time series using ARIMA. Forecating value will be 2 hours ahead value.

The following chart shows the performance of resample models in predicting the water level.

Although mean error is increase as expected, No significance difference in Variance of error