The increasing complexity of supply chains necessitates advanced risk prediction methods to mitigate disruptions and inefficiencies. Traditional risk assessment models often fail to capture sequential dependencies and evolving patterns in supply chain data. This study proposes a Bi-LSTM-based deep learning framework for supply chain risk prediction, leveraging bidirectional learning to enhance classification accuracy. The model is trained and evaluated on real-world supply chain transaction data, demonstrating superior performance over conventional machine learning and deep learning classifiers. Experimental results show that the proposed Bi-LSTM model achieves higher accuracy, recall, and F1-score, effectively identifying potential risks. By incorporating sequential dependencies, the model improves predictive reliability, addressing limitations of existing risk classification approaches. The findings highlight the potential of deep learning for robust supply chain risk management, paving the way for more adaptive and scalable predictive solutions.