SSO-FFNN: A Scalable Seeker Optimization-Enhanced Feedforward Neural Network for Predicting Anxiety Levels in College Students

Abstract

This study aims to predict anxiety among college students using optimization-enhanced neural network models instead of conventional machine learning. The student anxiety and depression dataset from Kaggle (6,982 records) was analyzed through comprehensive text preprocessing (tokenization, stop-word removal, TF-IDF feature extraction, and SMOTE balancing), exploratory data analysis, and a set of baseline models including ANN, RNN, ES-ANN + LOF, Harmony-Search ANN, and GA-BP ANN. To address the limitations of earlier approaches, a new SSO-FFNN (Scalable Seeker Optimization-Enhanced Feedforward Neural Network) framework was introduced, where the SSO algorithm optimizes network weights and biases to avoid local minima and improve convergence speed. Results show that the proposed SSO-FFNN achieved 97 % accuracy, an F1-score of 0.84, and a ROC-AUC of 0.97, outperforming all baseline models. These findings highlight the potential of optimization-driven neural networks for early anxiety detection and timely preventive counselling in academic settings.