<p>Wildfires pose a serious threat to environmental sustainability, economic infrastructure, and human safety. The accurate and timely classification of wildfire events is essential for effective mitigation and disaster response. This paper proposes a hybrid deep learning model that integrates a convolutional neural network (CNN) and a bidirectional long short-term memory (BiLSTM) network to enhance the prediction and classification of forest fire occurrences. The model is trained and evaluated using the Algerian Forest Fires Dataset, which contains 244 instances with key meteorological and fire-related parameters, including temperature, relative humidity (RH), wind speed (WS), rainfall, and fire weather indices such as the Fine Fuel Moisture Code (FFMC), Duff Moisture Code (DMC), Drought Code (DC), Initial Spread Index (ISI), Buildup Index (BUI), and Fire Weather Index (FWI). The proposed CNN-BiLSTM model effectively captures both spatial features and temporal dependencies within the data. It achieves a superior classification accuracy of 99.99%, a perfect area under the receiver operating characteristic curve (AUC-ROC) of 1000, and high performance across precision, recall, and F1-score metrics. Comparative analysis demonstrates the model’s superiority over other deep learning architectures, including Gated Recurrent Unit (GRU, accuracy: 93.88%, AUC: 0.966), and a baseline CNN (accuracy: 95.92%, AUC: 0.970). The evaluation metrics employed in this study include accuracy, precision, recall, F1-score, and the AUC-ROC. The outstanding results affirm the robustness and applicability of the CNN-BiLSTM model for early wildfire detection and classification. This paper contributes to environmental monitoring and risk assessment systems and sets the stage for future work involving the integration of additional ecological and topographical variables to improve model generalization across diverse forest regions.</p>