<p>Social networks increasingly influence the success of movies and other products, shaping public perception and decision-making. Businesses and individuals often mine data from social networks to gauge the public’s perception of a product, service, article, or event. Additionally, predicting movie popularity is crucial, as the film industry involves substantial financial risks. This study presents Deep Temporal Popularity (Deep-TemPo), a hybrid CNN–BiLSTM model designed to forecast weekly movie popularity using audience reviews collected from social media. Deep-TemPo extracts local n-gram patterns through multiple convolutional layers and captures sequential dependencies using a bidirectional LSTM, enabling the integration of both short-term textual features and temporal dynamics. The model incorporates a feature fusion layer to combine representations from multiple CNN components with the sequential features from the BiLSTM, producing a robust input for the prediction layer. Comparative experiments against alternative methods demonstrate that Deep-TemPo achieves an F1-score of 0.834, with a recall of 0.844 and a precision of 0.825, indicating its effectiveness for temporal popularity prediction in social media contexts.</p>