Accurate forecasting of Gross Domestic Product (GDP) is essential for informed economic planning and policymaking in Morocco. However, for many countries, the values of economic indicators are not reported every year, leading to missing data points. Numerous studies have focused on imputing these missing values, as well as predicting future variations of these indicators to support proactive decision-making. Among these indicators, GDP stands out as a critical measure of economic health, making it the focus of this paper. Traditional econometric models, like ARIMA and VAR, often fall short of capturing the intricate temporal and nonlinear patterns inherent in economic data. In this paper, we propose a novel deep learning approach for time series forecasting of Morocco’s GDP, using the ability of these models to capture dynamic trends and intricate relationships. Using historical GDP data from 1980 to 2010, Obtained from the Penn World Table and World Development Indicators, for training, and data from 2011 to 2019 for testing, we evaluate the performance of advanced deep learning models (LSTM, CNN, BD-LSTM, and ED-LSTM) and traditional econometric methods (ARIMA and VAR). Performance was assessed utilizing the Root Mean Squared Error (RMSE) metric. The results reveal that the BD-LSTM model significantly outperforms both traditional approaches and other deep learning models in terms of robustness and accuracy, particularly in the capture of short-term economic fluctuations. The BD-LSTM architecture includes both past and future contextual information, making it a more efficient model for GDP prediction and interpretation.