Predicting Consumer Preferences for Online Versus Offline Shopping Channels Based on Machine Learning Models

Le résumé fourni par la source

With the advent of the omnichannel retail era, identifying consumer shopping channel preferences has become critical for retailers to achieve precision marketing and enhance market competitiveness. Based on a consumer dataset from the Kaggle platform, which contains 11,420 valid samples and 18 predictive features, this study employed five machine learning models, namely Logistic Regression, Decision Tree, Support Vector Machine, Random Forest and XGBoost, to predict consumers' preferences for online versus offline shopping channels. The predictive performance of each model was systematically evaluated using metrics including recall, F1-score and AUC, and the key factors influencing consumers' channel choices were identified. Experimental results showed that Logistic Regression achieved the highest AUC (0.7147) and F1-score (27.34%), demonstrating the advantage of linear models in overall discriminative ability. XGBoost, with an F1-score (27.03%) close to that of Logistic Regression and a significantly higher recall (72.73%), exhibited more practically valuable comprehensive performance. In terms of feature importance, need for touch, tech-savviness, online payment trust and daily internet hours were identified as the four most critical features for identifying consumer channel preferences. This study reveals the core driving role of psychological and digital habits factors in channel selection, and provides decision support for omnichannel retailers to identify potential online consumers and formulate differentiated marketing strategies.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.