Shales are the common type of heterogeneity present in the oil sand reservoirs. Shale barriers usually have very low permeability leads to obstruction for fluid flow inside the reservoirs towards the wellbore. SAGD (steam assisted gravity drainage) and ES-SAGD (expanding solvent assisted gravity drainage) rely heavily on gravity drainage, and thus the impact of shales is more pronounced than the conventional recovery methods. Apart from reducing the oil drainage, shale barriers may also contribute to the heat losses. SAGD process is a well-known bitumen recovery technology. Hence, the impact of shale barriers is an essential aspect of understanding the bitumen recovery from oil sands. ES-SAGD is an emerging technology that combines solvent dilution effect along with steam to improve oil recovery while reducing carbon footprints. In this thesis, first, a procedure to construct the base 2D reservoir model is described. The procedure includes generation of log-normally distributed shale barrier realizations based on the mean and standard deviation. The impact of shale barriers on the recovery factor is described qualitatively using these realizations. Later, a non-linear model using a machine learning algorithm for predicting oil recovery factor using a synthetic dataset is being demonstrated. A detailed numerical simulation study has been conducted to analyse the impact of shale barriers on SAGD and ES-SAGD processes. Both recovery processes are analyzed over the whole simulation period. Additionally, they have also been evaluated at multiple cumulative steam-oil ratios (cSOR) cut-off at 2, 2.5, 3 and 3.5. Transition points are defined and explained to cluster the shale density/fractions based on similar behaviors. Co-injection of butane provided better recovery performance than hexane due to delayed condensation of C4 relative to C6.