Summary
In this chapter, we covered the first use application of forecasting in KNIME Analytics Platform; getting data from IoT sensors for modeling can be facilitated by KNIME workflows that have been loaded to KNIME Server and given REST endpoints. One workflow accepted this data and appended it to a database table until we had enough stored to train a model to generate humidity forecasts.
We introduced several classic methods for generating Time Series forecasts: the naïve forecast, which uses the most recent known value as its predictions; the mean value forecast that uses the mean of the known values as its predictions; exponential smoothing, which uses a weight average to generate its predictions, putting it somewhere in between the naïve and mean forecasts; and finally, the linear regression, the first model that is actually a fit. The ARIMA and SARIMA models later in this book will expand the regression format.
Finally, we talked about how to automatically...
