In the estimation of crude oil price prediction, parametric econometric and machine learning models are used, to predict the future price. The parametric econometric models include the conventional time series models by using the suitable(log-difference) transformation to fulfill the necessary assumptions according to the axioms of econometric modelling. The hybridization of ARIMA and GARCH is done to get the model with best predictions. The machine learning model (Recurrent Neural Network (Long Short-Term Memory) state of the art architect of neural network for sequential/time series data. The recent interest has been focused on developing the estimation technique to predict the future prices by using the series at level, rather than the return series. We find that the hybrid ARIMA-GARCH outperformed amongst all the models used in this study. But on theoretical basis and also the graph of predictions from RNN(LSTM) suggests that if we have to model high frequency data by estimating series at level rather than return series then one must go for the machine learning model RNN(LSTM). Supervisor:- Dr. Saud Ahmed Khan