A clustering approach for identification of profiles given barriers and facilitators for the practice of exercise during pandemic

Practising physical activity and avoiding sedentary lifestyle are essential for maintaining physical and mental well-being. Understanding the distribution of facilitators and barriers to physical activity among the population is essential for developing public policies that promote a more active lifestyle. Using data from a questionnaire applied during the COVID-19 pandemic, a clustering method was developed to identify individuals' profiles based on their barriers and facilitators to physical activity. The questionnaire comprised exclusively closed-ended questions, therefore, the proposed method operates with categorical datasets, applying Kmodes and ROCK clustering techniques, along with adaptations of silhouette score and Calinski-Harabasz evaluation techniques. The results obtained were validated and endorsed by health experts. The study revealed that barriers played a more significant role in determining consistent profiles. This research addresses a problem often overlooked in the machine learning literature: applying clustering algorithms to categorical data.