Stock Market Forecasting: Artificial Neural Network and Linear Regression Comparison in An Emerging Market
Year: 2005
Type: preprint
Abstract: In this research, we compare the forecasting performance of ANN and linear regression strategies in Istanbul Stock Exchange and we get some evidence of statistical and financial out perform of ANN models. Although the out-of-sample forecast accuracy statistics (RMSE, MAE and Theil's U) of ANN models which employ daily and monthly data are not better than the alternative regression models, we get significant evidence for a superior market direction prediction of ANNs. The ANN models correctly predict the signs of stock indexes up to 57,8 per cent, 67,1 per cent and 78,3 per cent for daily, weekly and monthly data respectively. The ANN models also generate better returns than the linear regression models when they are used as trading strategies. A hypothetical portfolio of 1 YTL initial value reach up to 2,76 YTL, 2,63 YTL and 3,35 YTL for daily, weekly and monthly data respectively, which are better than regression and buy-and-hold strategies.
Source: SSRN Electronic Journal
Authors: Erdinç Altay, Mehmet Hakan Satman
Institution: Istanbul University
Language: English
Field: Decision Sciences
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W2264104308