Output details
11 - Computer Science and Informatics
University of York
Testing Implications of the Adaptive Market Hypothesis via Computational Intelligence
<24>Adaptive and learning approaches are common in financial forecasting, yet their usefulness is questioned by a widespread assumption, The Efficient Market Hypothesis (EMH), which implies that markets follow a random walk. This paper identifies two properties of financial data, variable efficiency and cyclical profitability, that contradict the EMH claim. Showing that nonlinear dependence in a time series improves the efficiency of supervised machine learning, as confirmed on six different approaches, is an important methodological contribution to AI as a whole. This article won the best student paper award at CIFEr (2012), the main IEEE conference in this area.