Identifying Patterns of Substance Abuse in Tweets Using Deep Learning Model during Covid-19 (Preprint)

BACKGROUND Substance Use trend is highly seen during covid-19 due to various repercussion reasons amidst pandemic. Our study aims to make a comparison study a year before and after covid-19, analyze the themes and pattern associated with drug use in the study period. OBJECTIVE This work aimed to more accurately identify substance related posts, including references to specific drug types and utilization, and the purpose of drug use from large-scale social media data by training a deep learning model to monitor trends in temporal and spatial dimensions. METHODS Our method used self trained deep learning model on huge social media data to identify the post related to drug use followed by various statistical methods like k-means, LDA topic analysis, thematic analysis. RESULTS Our result showed that drug use increased dramatically by 20% just in 3 days of global pandemic declaration. Alcohol and cannabinoids remained the top discussed substances throughout the research period. Additionally, theme analysis highlighted the covid, mental health and economic stress as the leading issues that contributed to the influx of substance related posts during the study period. CONCLUSIONS This study highlights the trend of substance abuse during COVID-19 from the social media point of view. The results suggest that COVID-19 had a huge impact on mental health that corresponds to substance abuse, especially during the declared pandemic period.