Guided State Space (GSS)

GSS was developed in 2026 to replace TSD feature extraction in modeling. In extensive testing has produced better results with much less computational resources. It also supports reverse modeling where, starting with target features, it can determine the transformation necessary to optimize the utility of those features.

GSS transforms time series data into a single state space that is "guided" by a rule, such as maximizing entropy. About 200 features are then extraced from about 350 low pass bands in the state space or about 70 total "candiate" features of which a high percentage is usually statistically significant.