Time forecasting model algorithms for detecting trends and seasonal components in data

This algorithm is based on the identification of features in the data, and works in five steps:

1. Preprocesses data to correct anomalous behaviors and to prepare the time series for the analysis.
2. Identifies the most appropriate regression curve and removes the best trend from the data.
3. Analyzes detrended data to evaluate whether basic and detailed seasonal components are relevant.
4. Models the time series, deprived of trend and seasonal components, using autoregressive integrated moving average (ARIMA) techniques to discover serial correlations among samples.
5. Predicts future data using information collected in the earlier steps of the algorithm.

This algorithm is suitable for almost any type of time series, but it is particularly useful for data showing a clear model (regression, seasonality, or serial correlation) underlying the data.

This algorithm applies the #Automated Box and Jenkins algorithm to only the automatically detected last ramp of the data.

This algorithm is suitable when the time series shows a clear change of behavior in its latest samples, for example, a step or a slope switch.

BMC Capacity Optimization selects a set of available algorithms and then identifies the specific mathematical algorithm that achieves the best performance in predicting the given time series.

This is the default algorithm setting for time forecasting models. BMC recommends using it unless you have a specific reason to manually select a specific mathematical algorithm.