A cascade model of support vector regression and adaptive neuro-fuzzy inference system for next day stock price prediction

Abstract

Stock pricing is one of the challenging tasks in prediction due to noisy patterns with a slow changing curve. Global prediction techniques such as support vector (SV) show good enveloped prediction patterns but it tends to delay the prediction. Fuzzy prediction methods have better local optimizing and show significantly within training sets. Unfortunately, these sometimes generate surface oscillation effects in the output. This includes both global and local stock price rules with filtering of existing prediction models, output component base (OCB) and output-input iteration (OII) models, resulting in significant compromise for stock prediction.