Machine Learning Prediction of Anxiety Levels in the Society of Academicians During the Covid-19 Pandemic

Abstract

Various sectors in Indonesia have been impacted by the COVID-19 incident, such as the trade, health, entertainment, and social sectors. Although several steps have been taken to minimize the coronavirus's impact, problems still occur, especially in the education sector, which must carry out one of the challenges faced in the learning process during the pandemic. However, the environment and learning process that turned into distance learning caused the interaction with friends to decrease, and academics could only move in a limited space, making them overwhelmed by feelings of anxiety. Anxiety must be detected early and managed properly not to cause mental deterioration. Therefore, the researcher aims to predict academic anxiety based on the self-rating anxiety scale (SAS), demography, family, lifestyle, and employment using k-means. Furthermore, tested the prediction results obtained with a confusion matrix in accuracy, precision, and recall. The test results found the accuracy rate is 99%, precision is 98% (moderate level), 100% (normal level), and recall is 97% (normal level), 100% (moderate level). These results indicate that the k-means on demographic, family, lifestyle, employment, and SAS aspects provide optimal results for predicting the anxiety level of the BM University academic community.