دانلود کتاب هوش مصنوعی برای بازارهای مالی – رویکرد چند مدلی

This book exposes a manageable approach to exploit nonlinearity for the purpose of financial investment. It summarizes the research Raphael Douady and co-authors have conducted for nearly two decades now. This research is expanded on multiple aspects by a joint work conducted by Thomas Barrau. In general, relationships between variables are nonlinear and noisy. When attempting to relate a target variable to several explanatory variables, modeling finely multi-dimensional linearity suffers from the curse of dimensionality without bringing blatant benefits. Polymodels simplify this approach by retaining unidimensional nonlinearities and proposing various aggregation methods according to the task at hand. For instance, in the case of prediction, each one-dimensional model produces a prediction. The most relevant ones are then selected, then possibly averaged in case of ambiguity, to yield the final prediction. In the case of defining a risk indicator, as downturns are usually associated with an increase of correlations, the distribution of goodness-of-fit metrics is analyzed. It can be seen as a form of ensemble learning.