Financial Prediction using Inductive Models

Abstract

Financial prediction is an example of a prediction problem which is challenging due to small sample sizes, high noise, non-stationary, and non-linearity. Neural networks have been frequently used in financial prediction because of their ability to deal with uncertain, fuzzy, or insufficient data. Despite that, neural networks(NN) have limitations; they still require a significant amount of a priori information about the model structure. Group Method of Data Handling (GMDH) is an inductive approach which attempts to overcome the subjectiveness of neural networks based on the principle of self-organization. We have developed an algorithm inspired from the evolutionary manner of conventional GMDH to generate an inductive model based on using multilayer perceptron that can avoid some of GMDH problems like the exhaustive computations on candidate Adalines and the increasing number of Adalines in the following layers