Data Mining for Overreaction in Financial Markets


Duran A. , Caginalp G.

IASTED International Conference on Software Engineering and Applications (SEA), Arizona, United States Of America, 14 - 16 November 2004, vol.467, pp.28-35

  • Publication Type: Conference Paper / Full Text
  • Volume: 467
  • City: Arizona
  • Country: United States Of America
  • Page Numbers: pp.28-35

Abstract

We study overreaction and the cumulative effect of the consecutive local overreaction patterns in financial markets. The ”overreaction diamond” pattern [1] is one of the key components of a financial market bubble. The cumulative effect of the consecutive short term overreactions arising from the deviation of stock prices from their fundamentals can be explained by attribution theory, feedback traders, affect and representativeness theories, and reference points in investments. We study large set of financial data and propose a data mining method by exploiting the relative cumulative sentiment of the investors. This leads to a potential for the implementation of suitable algorithms and the preparation of software packages that can be useful for prediction of various stages of overreaction and bubbles.

Keywords: Data mining, overreaction, computational finance software, financial markets, and bubble.