An intelligent predictive framework for identifying mental state using facial expression data

Automatic facial expression analysis has gained significant attention for supporting intelligent mental state assessment systems. However, many existing deep learning–based facial expression recognition models rely solely on convolutional architectures, without integrating optimization strategies to enhance feature selection and generalization. To address this limitation, this study proposes an Ant Lion–based optimized GoogleNet framework (ALbGN) for mental state prediction from facial expression images. The proposed approach combines deep feature extraction using GoogleNet with Ant Lion Optimization (ALO) to improve feature selection and classification performance. The model is evaluated using the publicly available Facial Expression Recognition (FER) dataset obtained from Kaggle, comprising 35,887 grayscale facial images categorized into seven expression classes: Happy, neutral, angry, surprise, disgust, fear, sad. To ensure robustness and prevent performance bias, stratified 5-fold cross-validation is employed, preserving class distribution across folds. In each iteration, 80% of the data are used for training, and 20% for validation/testing, and final results are reported as the average performance across all folds. The framework is implemented in Python (version 3.7.14) and trained for 100 epochs per fold. Performance is assessed using accuracy, precision, recall, F1-score, and loss analysis. Experimental results demonstrate that the proposed ALbGN framework achieves consistently high classification performance across all folds, indicating strong generalization capability. The integration of Ant Lion Optimization enhances discriminative feature learning and improves classification stability compared to conventional deep learning approaches. The proposed hybrid framework provides an effective and scalable solution for automated facial expression–based mental state prediction, with potential applications in intelligent mental health monitoring systems.