With the integration of online and offline retail channels, consumers’ consumption behavior has shown obvious cross-channel characteristics, and accurately predicting consumer cross-channel consumption paths has become a key issue for retailers to optimize marketing strategies and improve user experience. Existing prediction methods often ignore the spatio-temporal correlation of consumer behavior and the heterogeneity of channel attributes, leading to insufficient prediction accuracy. To solve this problem, this paper proposes a Consumer Cross-Channel Consumption Path Prediction model based on Spatio-Temporal Graph Convolutional Network (ST-GCN-CCPP). First, we construct a spatio-temporal graph structure that integrates consumer nodes, channel nodes and time nodes, in which the spatial correlation is characterized by the geographical distance between offline channels and the interaction frequency between consumers and channels, and the temporal correlation is represented by the time interval of consumer consecutive consumption behaviors. Second, a dual-channel attention mechanism is designed to adaptively adjust the weights of spatial and temporal features, and the channel attribute embedding module is introduced to model the heterogeneity of different channels (such as online e-commerce, offline physical stores, and social commerce). Finally, the graph convolution layer and gated recurrent unit (GRU) are combined to capture the deep spatio-temporal dependencies in the consumption path. Experimental results based on two real datasets (Retail Dataset and E-commerce Dataset) show that the proposed ST-GCN-CCPP model outperforms traditional methods (such as Markov Chain, LSTM) and existing graph neural network models (such as GCN, ST-GCN) in terms of prediction accuracy, precision, recall and F1-score. Specifically, compared with the best baseline model, the F1-score of ST-GCN-CCPP is improved by 8.3% and 6.7% on the two datasets respectively. This study provides a new technical framework for consumer cross-channel consumption path prediction and has important theoretical value and practical significance for retail enterprise marketing decision-making.
Xiaojun Cui (Sun,) studied this question.