Cross-border e-commerce platform for commodity automatic pricing model based on deep learning

With the innovation of information technology and the rise of the Internet economy, cross-border e-commerce has grown up to be an important means and strategy for enterprises to seek rapid development. This paper proposes a model that fuses CNN (Convolutional Neural Network) and attention mechanism to encode image features, and selects the image features of commodities. A 5-layer CNN without a fully connected layer is constructed to initially extract image features, and then a set of attention mechanism strategies is designed. This strategy is used to select the image features that have the greatest impact when generating words at different times. Considering the characteristics of quantitative indicators of the pricing model, this paper transforms this evaluation process of consumers into price perception. Corresponding mathematical model is set up to improve and expand the original probability unit model. The consumer selection model is utilized to obtain a prediction of product market share, and a nonlinear constraint programming is established to determine the optimal price. The strategy takes into account the changed market shares of consumer characteristics and product quality evaluation results. In the two-layer hybrid channel supply chain model, retailers and manufacturers all use third-party platforms when they achieve maximum benefits; when price cross-elasticity coefficients and third-party platform usage fees are independent variables of influencing factors, retailers are dispersed on CNN to get the most profit under the pricing strategy. Similarly, when the unit product tax difference is the independent variable of the influencing factors, the manufacturer is also the most profitable under the CNN decentralized pricing strategy.