Predicting Stock Index Volatility Using Artificial Neural Networks: An empirical study of the OMXS30, FTSE100 & S&P/ASX200

Sammanfattning: In this thesis I study the performances of artificial neural networks (ANNs) and three various ARCH-type models to predict weekly volatility of the Swedish (OMXS30), the British (FTSE100) and the Australian (S&P/ASX200) major stock indices. The three various ARCH-type models are the GARCH(1,1), the EGARCH(1,1) and the TGARCH(1,1). The purpose is to investigate if ANNs outperform the more traditional ARCH-type models in predicting weekly stock index volatility. An out-of-sample testing methodology is applied to the most recent 20 percent of the data observations, which fully range from 8th February 2008 to 29th December 2017. The metrics used to evaluate the volatility-predicting performances of the different models are the RMSE, the MAE, the MAPE and the out-of-sample sample R^2. The results show no evidence of ANN predicting superiority for any of the three stock indices.