This research proposes a fashion recommendation system (FRS) utilizing a small amount of zero-party data sourced from self-reported consumer profiles featuring their psychographic, demographic and physical characteristics that influence their fashion item purchasing decisions. The aim is to ascertain whether various antecedent variables influencing fashion purchase decisions can be applied to the FRS from an integrated perspective. The obtained findings verify the feasibility of developing an FRS using decision trees based on consumer zero-party data.
A decision tree model was constructed to predict consumer design preferences for fashion socks. The respondents’ self-reported data (i.e. consumers’ zero-party data) were used to estimate the relevance of this information for their fashion item purchase decisions based on the Classification and Regression Trees (CART) model.
The findings indicated that the use of consumers’ zero-party data is an effective way to improve the FRS. Specifically, this research identified the top 20 features that play a crucial role in prediction and demonstrated high CART model accuracy.
This study examined the feasibility of an FRS using a small amount of data comprising consumers’ psychographic, demographic and physical features. This interdisciplinary investigation expands the scope and depth of research on recommendation systems. This new approach can also help fashion companies (especially small- and medium-sized enterprises) build proprietary recommendation systems.
