Support Vector Regression with Levy Distribution Kernel for Stock Forecasting

Abstract

-Normal distribution has been widely applied in modern financial time series forecast. Lé vy distribution is another alternative to normal distribution which this paper would like to explore. It has been demonstrated in this paper that Support Vector Regression (SVR) using Lé vy distribution kernel is a robust forecasting tool and performs very well in the following experiments. Three stock Indexes are selected to test the (SVR) forecasting model. They are Hong Kong -Hang Sang Index (HSI), U.S. -Dow Jones Industrial Average Index (DJ) and China -Shanghai Composite Index (SH). It has been discovered that the SVR using Lé vy kernel has given better performance in 9 out of 24 tests when compared with that of the commonly used RBF kernel. It shows promising result in general.