Emotion recognition plays an important role in affective computing and human–computer interaction. Previous research has used machine learning techniques to cluster emotions using the Pleasure–Arousal–Dominance (PAD) emotional model. In the existing approach, clustering methods such as K-means and machine learning algorithms like Decision Trees and Support Vector Machines were used to classify emotions based on PAD scores. However, the approach suffers from limitations such as overfitting, limited datasets, and inability to detect mixed emotions. This paper proposes an enhanced emotion recognition framework that integrates deep learning models with hybrid clustering techniques to improve classification accuracy and generalization. The proposed system uses dimensionality reduction, advanced clustering, and neural network–based classifiers to better map PAD scores to emotional categories. Experimental analysis shows that the proposed system improves emotion classification performance and provides a more scalable framework for real-world affective computing applications.