A Novel Method for Predicting Consumer Purchase Behaviour on E-Commerce Platforms through Graph Neural Network Model

Men and women display divergent behaviours influenced by variations in development, upbringing, and an array of social, biological, and environmental factors. These gender-based disparities substantially influence both personal and professional choices, encompassing dietary preferences and consumer purchasing behaviour. Disparities arise along the decision-making process, from requirement identification to option evaluation and selection, underscoring the importance of tailored approaches. This work presents a data-driven approach that incorporates advanced preprocessing, feature selection, and model training to examine gender-based behavioural trends. Preprocessing involved min-max normalisation and the transformation of numerical data into categorical values. Permutation importance and SHAP values were employed to discern critical aspects affecting decisions. A Degree-free GNN was utilised to model behavioural patterns. The model attained an accuracy of 96.27%, surpassing conventional methods such as LightGBM, XGB and RF in detecting gender-specific distinctions. The findings emphasise the substantial influence of gender on consumer purchasing behaviour and decision-making processes. These findings endorse the implementation of customised, gender-sensitive marketing and recommendation systems. The study provides a comprehensive framework for comprehending and forecasting gender-specific behavioural inclinations in consumer decision-making scenarios, with possible applications in targeted marketing and customised user experiences.