Explanation of binarized tick data using investor sentiment and genetic learning

Abstract

This paper attempts to clarify some time series properties of binarized tick data by investor sentiment and genetic algorithm. For this purpose, first we explore the conditions for genetic algorithm to describe investor sentiment. Then we calculate auto-correlations and conditional probabilities using binarized sample paths generated by estimated models of investor sentiment. The most fitted parameter set of genetic algorithm have the following implications: First, a herd behavior is likely to emerge. Second, traders try to perceive brand-new information even if it is not completely correct.