Predicting Sustainable E-Commerce Delivery Preferences Using Emotional Neural Networks: A Smart City Logistics Perspective

There is growing interest among researchers and practitioners in understanding and predicting online consumer behaviour, particularly regarding environmental considerations in e-commerce and sustainable urban logistics. This study examines consumer attitudes and preferences and predicts the acceptance of sustainable delivery options in an emerging market context. Drawing on 355 valid survey responses from Serbia, an Emotional Neural Network (ENN) framework was adapted from its original regression formulation to binary and multiclass classification of consumers’ willingness to accept sustainable delivery options based on demographic and sustainability-related attitudinal variables. SHAP analysis was used to examine the contribution of individual predictors, while descriptive findings provided additional context on consumer responses to ecological, financial, social, and combined sustainability incentives. The results indicate broad support for sustainable delivery options, especially when linked to transparent incentives such as lower costs, loyalty points, or charitable contributions. Models based on attitudinal variables achieved a macro F1-score of about 0.63 and a ROC-AUC of about 0.69 in binary classification, comparable to logistic regression and random forest, whereas demographic variables alone performed close to chance level. Green buyer status and social and environmental attitudes contributed most to the predictions. The findings offer exploratory insight into stated sustainable delivery preferences within this sample. Their relevance to smart city logistics requires validation using observed behaviour and operational data.