Accurate short-term electricity price forecasting is a key input in the operational scheduling decisions made by electricity market participants. Thus, there has been a board body of models and techniques presented in the literature to forecast these prices. However, when it comes to real-time markets, they are significantly more volatile than day-ahead markets. This is because they are affected by unforeseen real-time grid events that may lead prices to jump to unusual high/low levels. This high volatility makes accurate forecasting of electricity prices a very challenging task. In this thesis, a volatility analysis of the Alberta electricity market is carried out to demonstrate significant intra-day fluctuations in the price patterns. Relevant publicly available data from the Alberta electricity market is investigated to select a set of explanatory real-time variables that might help predict some of these fluctuations. The Long Short-Term Memory network combined with the XGBoost decision tree algorithm is applied in a multi-stage forecasting mechanism to generate forecasts up to 96 hours ahead. Experimental results demonstrated that the proposed model provides a better performance for short-term electricity price forecasting relative to existing models. Two economic applications of the developed forecasting model are carried out to demonstrate its usefulness to the market players of the Alberta electricity market. The results obtained from these applications show that, compared with other approaches, the deployment of the developed forecasting model can increase significantly the revenues of several market players of the Alberta electricity market.