FORECASTING ELECTRICITY PRICES IN CANADA: A COMPARATIVE ANALYSIS OF ARIMA, LSTM, AND XGBOOST MODELS FOR FINANCIAL DECISION-MAKING

Abstract Forecasting electricity prices is important for smart decisions in the energy field. This includes investors, utility companies, and students studying energy finance. In Canada, more provinces are gaining control over their power systems and using more renewable energy. These changes have made electricity prices more unstable and harder to predict. This study compares how well three models can forecast prices: Autoregressive Integrated Moving Average (ARIMA), Long Short-Term Memory (LSTM) networks, and Extreme Gradient Boosting (XGBoost). The models are tested using hourly electricity price data from Alberta, taken from the Alberta Electric System Operator (AESO) from 2015 to 2023. To measure how accurate the models are, we use three common metrics: Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and Mean Absolute Percentage Error (MAPE). The results show that ARIMA works fairly well with steady data but performs poorly when prices change quickly. LSTM does better by recognizing patterns over time. XGBoost gives the best results overall, especially when handling complex and changing price trends. Although none of the models is perfect, machine learning models like XGBoost seem to be more reliable. These findings can help guide future research and support better forecasting in Canada’s electricity market.