Detecting noise trading using fuzzy exception learning

Th paper analyses noise trading, a phenomenon observed in financial markets when technical traders forecast financial price movements based on recent prices and volumes. Due to noise trading, financial returns show (during a few and unknown periods in time), certain deterministic behavior besides the usual random behavior predicted by the efficient market hypothesis. Our goal is to unmask the (fuzzy) deterministic part, that is, to discover the special circumstances called ´regimes´, under which noise trading takes place. To reach our goal, we use the Competitive Fuzzy Exception Learning Algorithm (CELA) as introduced by W.M. van den Bergh, and J. van den Berg (2000), J. van den Berg, and W.M. van den Berg, (2000). In order to analyze the properties of our method, we apply it on an artificially made, yet quite hard to analyze financial time series. Even in a very general setting, CELA appears to be able to discover various important ´regimes´ corresponding to exceptional price developments. These occurrences are collected in a fuzzy rule base