I am reading the book, "Machine Learning in Java" published by Packt publishing, and written by Boštjan Kaluža; and there is a paragraph at the book, and I can not understand.
At the book, there is a concrete dataset that comprises real traffic to Yahoo services, along with some synthetic data. Here is the snippet of the data:
The author wants to detect anormaly in this dataset, and he wrote the paragraph below:
Detecting anomalies in raw, streaming time series data requires some data transformations. The most obvious way is to select a time window and sample time series with fixed length. In the next step, we want to compare a new time series to our previously collected set to detect if something is out of the ordinary. The comparison can be done with various techniques, as follows:
Forecasting the most probable following value, as well as confidence intervals (for example, Holt-Winters exponential smoothing). If a new value is out of forecasted confidence interval, it is considered anomalous.
He hasn't mentioned how to "forecast the most probable following value" afterwards. How can one do it in concrete terms?