In this study, the researchers looked at which behavioural and monetary tracking features in online gambling and which machine learning algorithms best predict self-reported problem gambling. The researchers analyzed data from 1,661 people who gambled on a North American gambling website. They used a mix of self-reported data on problem gambling and data from participants’ online gambling accounts. Behavioural tracking features, like number of bets made and time spent gambling, predicted self-reported problem gambling better than monetary tracking features, like amounts of money won and lost. Logistic regression and random forest were the machine learning models that best predicted problem gambling. Participants experiencing problem gambling were younger in age, more likely to be male, bet more money, bet more money on casino games and less on lotteries, deposited more money into their online accounts, had more failed deposits, and gambled for longer sessions.The findings can inform responsible gambling strategies and interventions that gambling operators can implement to reduce harm from gambling among users.
