The proliferation of social media has led to a surge in social bots that manipulate public opinion and spread misinformation. To address this issue, we propose a novel deep learning model for social bot detection based on neural network technique by considering the temporal features in social network data. Our model leverages temporal, linguistic, and social network features by incorporating temporal embeddings, word embeddings, and social embeddings. By capturing the dynamic patterns of user behavior over time, along with linguistic and social network characteristics, our model can effectively distinguish between human users and bots. Experimental results on a large-scale Facebook dataset demonstrate that our model achieves a remarkable accuracy of 0.943 in detecting social bots, highlighting the effectiveness of multi-modal data fusion in enhancing detection performance.