Behavioral Cluster Analysis of Online Gambling and Its Association with Big Five Personality Profiles
The swift development of online gambling has posed new issues in determining risky behaviours of players and psychological factors that relate to gambling behaviour. In contrast to the traditional gambling environment, the online gambling environment is always able to record behaviour which means that behavioural patterns can be analysed using data analytics. This thesis explores how online gambling behaviours are aggregated in groups of players, and how the behaviours are associated with the big five personality traits. The paper examines big data of online gambling through machine learning methods to uncover latent behavioural groups, where the frequency of gambling, size of bets, degree of involvement, reward pursuit, losses and control over gambling behaviour are considered. Players were clustered into segments representing various gambling behaviours, including regulated and low-risk players, to very active, loss-oriented and impulsive gamblers, using clustering algorithms. There were five primary clusters of behaviour, including regulated low-risk players, highly engaged regular players, high-loss players and low-control unstable players. Simultaneously, the data about personality (according to the Big Five model (Openness, Conscientiousness, Extraversion, Agreeableness, Neuroticism)) were analysed to create five personality clusters. The behavioural clusters were compared to personality clusters using matching based on similarities. The results indicated that there is very high structural similarity between the two. The group of individuals exhibiting erratic, risky and excessive gambling behaviours were usually linked to the high levels of neuroticism and low levels of conscientiousness and moderate and controlled gambling behaviours were more linked to the emotionally stable, conscientious and responsible individuals. Results suggest that the type of gambling behaviours are not by chance, but they can be an expression of more general psychological dispositions. The study also looks at prediction models to identify more high-risk gaming behaviours by use of simulated behavioural characteristics. Machine learning models, such as K-Nearest Neighbors (KNN), Support Vector Machine (SVM), Random Forest (RF), Extra Trees, Gradient Boosting (GB), Logistic Regression (LR), Adaptive Boosting (AdaBoost), and Decision Tree (DT) were applied to predict high-risk and low-risk gambling behavior. Among the tested models, Logistic Regression achieved the best overall performance, with a test accuracy of 99.04% and an F1 score of 98.81% This indicates the excellent predictive power of behavioural attributes and effectiveness of behavioural analytics in automated monitoring. This paper demonstrates the feasibility of reinterpreting online gambling behaviours into useful behavioural clues, and associating them with personalitybased behavioural patterns in a structural manner. The proposed paradigm can help to improve the emergence of gambling technologies to support responsible gambling, early detection systems and future interdisciplinary studies including machine learning, psychology and online behavioural analytics.