A novel hybrid machine learning model for shopping trip estimation: A case study of Tehran, Iran

Abstract

Online and offline shopping trips have a profound impact on various facets of urban life, including e-commerce, transportation systems, and sustainability. To assess the factors shaping consumers' decisions, we introduce a novel hybrid machine learning model that integrates the Gray Wolf Optimization (GWO) algorithm with a deep Convolutional Neural Network (CNN). This model is applied to predict shopping behavior based on a survey of 1000 active e-commerce users residing in areas 2 and 5 of Tehran. These individuals have made successful purchases through both online and offline services during the final 20 days of 2021. The GWO algorithm plays a pivotal role in selecting optimal features and hyperparameters for the deep Convolutional Neural Network, which is a powerful deep learning model for image recognition and classification. Notably, our model achieves an impressive accuracy of 97.81% while maintaining a MSE of 0.325, having identified seven out of ten key features as the most influential. To gage the effectiveness of our approach, we conduct a comparative analysis with alternative methods. The results unequivocally showcase the superiority of our proposed algorithm, which attains an accuracy of 97.81%. In contrast, other models such as CNN, LSTM, MLP, DT, and KNN yield accuracies of 95.63%, 94.04%, 90.12%, 86.49%, and 80.16%, respectively. This study offers valuable insights for transportation planners, e-commerce managers, and policymakers. Its primary objective is to assist them in formulating effective strategies aimed at reducing transportation costs, curbing pollutant emissions, mitigating urban traffic congestion, and enhancing user satisfaction all while fostering sustainable development.