This study proposes a hybrid deep learning model that combines Bidirectional Long Short-Term Memory (Bi-LSTM) networks with Transformer-style attention mechanisms and technical analysis indicators to forecast foreign exchange (Forex) rates involving the Albanian lek (ALL). Using daily historical data of EUR/ALL (1999–2025) and USD/ALL (1991–2025) the model adopts a five-class classification system to capture the granularity of price movements. Additionally, historical macroeconomic events are embedded into the dataset to improve model contextual awareness. The forecasting performance of Bi-LSMT model is evaluated against traditional technical analysis methods to assess the feasibility and practical relevance for individual investors. Additionally, a hybrid forecasting framework is introduced, combining Bi-LSTM neural network with technical indicators to improve predictive performance. Experimental simulations, conducted using Python and real forex price data, demonstrate the model’s applicability and relevance for financial forecasting in emerging markets, highlighting its potential as a decision support tool for investors and policymakers.