Optimized feature-based image sentiment analysis using refined gated mechanism in GRU neural network model

The given manuscript introduces a new method for automated sentiment prediction using images called Optimized Refined-Gated Recurrent Units (OR-GRU). The proposed model uses GRU’s to extract spatial information and sequential patterns from face images, with an improved gating mechanism focusing on significant features for interpretability. Sine Cosine Algorithm (SCA) is used to optimize the retrieved characteristics, enhancing predictions' accuracy. The model outperforms traditional Convolutional Neural Networks (CNN)’s and other recurrent architectures used in previous studies in terms of accuracy (%) and root mean square error (RMSE). It is seen that the proposed model attains the highest accuracy (98.91%) and the lowest RMSE (0.28) as compared to conventional architectures used in the recent studies.