Tuesday, May 12, 2009

MTFS : a fuzzy approach


As far as MTFS is concerned, it is possible to implement some Fuzzy Behaviour Analysis on our indicator.

For simplicity, we will limit our analysis to the fast white line (WL) and the slow green line (GL).

We can now therefore implement a Fuzzy Logic Controller (such as our sFLC3) onto both MTFS lines in a very classical way as this exercise is somewhat akin to the typical thermostat already described on the ForeTrade web site (see reference below). We do indeed feed each FLC with level and gradient information, to extract a line behaviour as an output.

Note however that we will derive the fuzzy system into 8 variations according to 2 extra pieces of information, namely, according to whether lines have crossed and how (no or outdated crossover, early crossover, spot-on, and late crossover) and whether the WL has peaked or not (crisp boolean). The rule sets will obviously be incompletely defined as some fuzzy states are impossible or of no interest.

We shall also pay attention to state saturation when interconnecting the 2 FLCs.

Next step will be to link the MTFS behaviour to Entropy and Support/Resistance information.


Reference: http://www.seattlerobotics.org/encoder/mar98/fuz/flindex.html