Psychological and Mental Health Evaluation of English Language Students using Recurrent Neural Networks

Psychological health is crucial in educational settings and recognized as a significant feature in structuring behavior of teachers and learning outcomes of students. Timely and accurate identification of mental health issues aids in early intervention and initiation of the recovery process. Traditional assessment methods are subjective, time-consuming and faced different challenges. This study uses Recurrent Neural Networks (RNNs) to evaluate the psychological condition of students of English language. RNNs uses Long Short-Term Memory (LSTM) layers to capture long-term dependencies in language and develop a robust and efficient model that assesses students' psychological well-being through their written and spoken English. The RNN architecture is composed of several components. Firstly, it has an embedding layer that converts words into dense vectors of fixed size. Next, two stacked LSTM layers process these vectors and capture contextual information from the sequences followed by fully connected dense layers which transform LSTM outputs into psychological health scores. Finally, a sigmoid activation function in the output layer classifies the psychological state such as signs of stress or no stress. The data for this study includes essays, classroom discussions and interactions from English language learners. The data is preprocessed with tokenization, lemmatization and removal of stop words. To demonstrate the performance of RNN in forecasting English language student's mental health it is compared with different state of the art algorithms like Support Vector Machine (SVM), Artificial Neural Networks (ANN) and Random Forests (RF) in terms of accuracy, precision, recall, and F1-score. The results show high accuracy in predicting stress, anxiety and motivation levels outperforming its predecessors and leading to better teaching strategies and improved learning outcomes.