Modeling backward-looking volatility is of utmost significance to hedgeand price an option. Previous research has compared implied volatility’sability to forecast future movements to historical volatility.In this work, the contribution to the literature is mainly focusingon Long short-term memory (LSTM) neural networks predictabilityfor backward-looking volatility forecasting of the Indian S&PCNX Nifty. Using daily SPX options data spanned over the periodfrom 2/5/2001 to 26/2/2020, we find that LSTM fits the trainingdata exceptionally well, with a coefficient value of 0.92 in a singleregression evaluation model and coefficient value more than 0.9 inall multiple regression models. From the empirical analysis, we findthat LSTM outperforms three other state-of-the-art volatility forecastingmodels in predicting the volatility in the out-of-sample dataset