Dynamic Time Warping to capture dynamic cycle shifts – Apr 2021
Combining Dynamic Time Warping with classic cycle analysis to capture the dynamic nature of cycles
The key in projecting cycles into the future is to capture the dynamic component of phase and length shifts as exact as possible. Otherwise all future projections will fail as cycles are never as symmetric as the sine math functions suggests. The DTW (Dynamic Time Warping) framework is another promising technique which can help us in selecting the set of cycles from the cycle scanner which did capture the dynamic shifts best in the past to build a composite forecast.
This is a research project demonstration which gives some insights in my research process. The outcome will lead to an optimized new feature called “Dynamic Auto Composite”. This new feature will select an auto-composite for you based on the dynamic procedure shown in this session.
