Prediction Model for Depression Associated with Online Learning during Pandemic using K-means Clustering and Particle Swarm Optimization

Most prediction models for depression do not focus on online learning as a primary area of concern. Hence, the study utilized data mining techniques and machine learning to design a prediction model that could forecast the factors related to depression in online learning. Moreover, the model used dataset normalization, attribute selection, clustering, and optimization techniques such as K-means and particle swarm optimization to extract the predictors from the data. The model identified 18 predictors out of the 55 predictors listed in the following instruments used in the study: the Patient Health Questionnaire-9, the University Student Depression Inventory, and the Quality of Student Life since Transition from a dataset of 305 respondents for the simulation. As a result, the model identified at least 28 respondents (9%) indicating depressive symptoms based on the data used to simulate the model's performance. Keywords : depression, clustering, COVID-19, prediction model, e-learning, optimization