LINEAR TREND DISCOVERY IN STOCK RETURNS USING ARTIFICIAL INTELLIGENCE CASE-BASED METHOD

Abstract

Relative values are reliable stock comparison tool. A stock value linear pattern was created, basing on one session return rate. Adaptation of Case-Based Reasoning method was developed for discovering sequences of session return rates, most similar to the linear pattern. The used local and global similarity functions were described. Empirical data covered 1130 close values of a Polish stock market main telecomunication company share. Two to seven session long session sequences, with highest similarity to linear model pattern with given increment value were discovered in the calculations.