Option pricing with neural networks vs. Black-Scholes under different volatility forecasting approaches for BIST 30 index options

This study compares the performances of neural network and Black-Scholes models in pricing BIST30 (Borsa Istanbul) index call and put options with different volatility forecasting approaches. Since the volatility is the key parameter in pricing options, GARCH (Generalized Autoregressive Conditional Heteroskedasticity), implied volatility, historical volatility, and implied volatility index (VBI) are used to determine the best volatility approach for pricing options according to moneyness and time-to-maturity dimensions. The paper also includes a subsample analysis in which the pricing performance of the models are evaluated during the turbulent periods. Overall results indicate that neural network outperforms Black-Scholes during tranquil times while Black-Scholes outperforms neural network during turbulent periods for call options. For put options, the Black-Scholes model is the best model during tranquil periods while neural network is the best model during turbulent periods.