Detection of dynamic cycles in financial data with a genetic algorithm
TradersWorld Magazine Issue 56
Cycle forecasts have been traditionally made based on the current active cycle, where the detected dominant cycle is considered static and extrapolated into the future. However, this assumption oversimplifies the behavior of the market and often results in poorly estimated future cycles without consideration of the dynamic components. Thus, a successful cycle-based trading approach should allow the user to follow the dynamic component of the dominant cycle and adjust the cycle forecast continuously, similar to the way in which a geographic positioning system (GPS) continuously adjusts the projected arrival time based on current traffic conditions.
Genetic algorithms (GAs) could provide such an approach by tracking market conditions and adapting parameters dynamically over time based on the underlying dataset. The GA approach differs from traditional digital signal processing in that, instead of attempting to evaluate all possible combinations, it uses a process of natural evolution to determine optimal solutions. […]