ADAPTIVE METHODS FOR FINANCIAL DECISIONS

This paper compares statistical techniques for trading equities in financial markets. Basic approach is identification of turning points of stock values by means of varying parameters and prediction errors of Markovian models. Recursive weighted estimators are fundamental tools for obtaining the necessary indicators and detecting changing points sequentially. The tuning coefficients of recursive algorithms can be selected either with statistical criteria or by maximizing the empirical gain on past data. An extensive application to the daily series of the leading US financial index is developed throughout to illustrate and evaluate the various methods. Experimental results are encouraging.