Predictive analytics for tobacco demarketing: a decision tree approach to consumer segmentation in Palestine

Purpose This study aims to create a decision tree-based segmentation model to classify consumers in Palestine as smokers or non-smokers to design demarketing strategies and optimize policy interventions in a resource-scarce setting. Design/methodology/approach The study used a cross-sectional design with respondents recruited via social media through convenience sampling. The final sample was 700. The study used listwise deletion to address missing data. The researcher trained a decision tree classifier with a split of 70:30, Gini Impurity as the split criterion and minimum cost-complexity pruning to address overfitting. Model performance was assessed using accuracy, precision, recall and the F1 score. Findings The model classified demographic and policy exposure data. This included age, gender, social networks, exposure to anti-smoking policies and peer networks. The study found males of young age with peer networks were most likely to smoke. The model achieved classification with high performance. A high-scoring, young, separated male, socially networked, was identified as a target for demarketing and resource intervention. Research limitations/implications The study assumes no causality. Deficiencies of the study include lack of evidence, self-based survey, one world and new variables. Future studies should aim to encompass more variables and greater timing. Practical implications Explanation of demarketing based on tobacco availability creates a challenge for peer smoking and socially influenced environments. Originality/value This is the first Palestinian research combining public health and machine learning (decision trees) to predict smoking behavior. It is among the first studies in unstable, lower-middle-income contexts and considers the socio-economic and political conditions of the West Bank. The study advocates behavior-specific, data-driven anti-tobacco policies. Industries use the marketing of smoking as a behavior that can be reduced using demarketing.