Gradient One-to-One Optimizer and Deep Learning based Student Stress Level Prediction Model

Student stress-based issues are considered as the most common reason in the student environment. Student stress level prediction is the major source for students’ academic performance and health. Students' stress levels increase the prevalence of psychological as well as physical challenges like nervousness, anxiety, and depression. Over the past years, different machine learning and deep learning based models have been proposed for student stress level prediction but they suffer certain limitations such as complex structure, less efficiency, high chance of misclassification, high chance of making mistakes. Predicting stress levels at early stage may help to minimize its impact and various serious health problems pertaining to this mental state. For this, automated frameworks are needed to predict stress levels accurately. This study proposes a hybrid approach named as GOOBO: DSNN (Gradient One-to-One Based Optimization: Deep Spiking Neural Network), that may identify stress accurately and efficiently utilizing optimization based hybrid of deep learning techniques. Here, the GOOBO is designed by incorporating Stochastic Gradient Descent (SGD) and One-to-One Based Optimization (OOBO). Here DSNN has been used which uses spiking neurons having different learning dynamics compared to traditional artificial neurons. Here proposed stress prediction model’s effectiveness has been enhanced by bio-inspired nature of DSNN simulating biological neural systems. The performance of the proposed GOOBO-DSNN is analyzed for its effectiveness using evaluation metrics such as accuracy, sensitivity, specificity, and precision. The proposed GOOBO-DSNN attained the maximum accuracy, sensitivity, specificity, and precision as compared to recently developed models. The proposed GOOBO-DSNN accomplished the higher accuracy, sensitivity, specificity, and precision of 90.976 %, 91.698 %, 91.336 %, and 90.179 % respectively. Duplicate attributes have been deleted, and missing values are filled in during the preprocessing step of the dataset.