Multimodal sentiment analysis is automatically detecting, analysing and extracting emotions and opinions in multimodal data. Deep learning is becoming very popular for multimodal sentiment analysis because it can automatically extract meaningful and abstract semantic features. The main objective of this study was to introduce an efficient model for multimodal sentiment analysis using deep learning method. The system is divided into four parts namely data layer, single-modality feature extraction layer, multimodal features fusion layer and sentiment analysis layer, it adopted the public dataset: CMU-MOSI and CMU-MOSEI. It can be concluded that the system can improve and surpass the traditional textual sentiment analysis