Given the expanding high-level opening to the outside world, there is a potential correlation between consumers’ behavioral patterns on cross-border e-commerce platforms and economic policies. This paper proposes a model for extracting consumer behavioral willingness features based on a convolutional neural network with sentence vector improvements, which optimizes the effect of cluster analysis on consumer behavioral willingness. The extraction model is utilized to extract consumer behavioral willingness features, and the outcomes demonstrate that the proportion of consumption and recommendation willingness is 47.65% and 52.05%, respectively, with a positive clustering effect. The regression results show that consumers’ behavioral willingness to cross-border e-commerce platforms increases with their preference for free trade and mixed trade policies, with an impact coefficient of 0.095 and 0.096. The degree of free trade simultaneously influences consumers’ willingness to recommend, with an impact coefficient of 0.154. However, the import substitution policy has an impact coefficient of −0.078, indicating that consumers are less likely to recommend cross-border e-commerce platforms to others.