Utilizing Pattern Mining and Classification Algorithms to Identify Risk for Anxiety and Depression in the LGBTQ+ Community During the COVID-19 Pandemic

Abstract

In this paper, we examine the results of pattern mining and decision trees applied to a dataset of survey responses about life for individuals in the LGBTQ+ community during COVID, which have the potential to be used as a tool to identify those at risk for anxiety and depression. The world was immensely affected by the pandemic in 2020 through 2022, and our study attempts to use the data from this period to analyze the impact on anxiety and depression. First, we used the FP-growth algorithm for frequent pattern mining, which finds groups of items that frequently occur together, and utilized the resulting patterns and measures to determine which features were significant when inspecting anxiety and depression. Then, we trained a decision tree with the selected features to classify if a person has anxiety or depression. The resulting decision trees can be used to identify individuals at risk for these conditions. From our results, we also identified specific risk factors that helped predict whether an individual was likely to experience anxiety and/or depression, such as satisfaction with their sex life, cutting meals, and worries of healthcare discrimination due to their gender identity or sexual orientation.