Computational models for short-term prediction of the stock market

Article type: Research Article

Authors: Menconi, Lorenzo | Gori, Marco | Lippi, Marco

Affiliations: Dipartimento di Ingegneria dell'Informazione, Università degli Studi di Siena, Siena, Italy

Note: [] Corresponding author: Lorenzo Menconi, Dipartimento di Ingegneria dell'Informazione, Università degli Studi di Siena, Siena, Italy. E-mails: [email protected] (L. Menconi); [email protected] (M. Gori); [email protected] (M. Lippi).

Abstract: The pervasive role of computers in finance is strongly affecting the way investors operate on the stock market. In the last few years there has been a strong focus on short-time transactions that is the subject of many studies in the emerging field of computational finance. In this survey, we discuss the nature of the problem and give an overview of the most significant results. The critical analysis given in the paper covers classic time series predictions techniques, with emphasis on methods widely used in AI, as well as more recent attempts to make predictions on the basis of textual information extracted from financial portals and other relevant on-line sources. We claim that most results in the literature are not based on systematic evaluation over long time intervals and that, at the same time, vague declarations on the supposed efficiency of the market are not motivated.

Keywords: Computational models, computational finance, predictions techniques

DOI: 10.3233/IA-2011-0025

Journal: Intelligenza Artificiale, vol. 5, no. 2, pp. 217-227, 2011