The development of international trade makes the fluctuation of commodity retail prices increasingly complex. Traditional commodity price forecasting methods are often based on statistical models and linear regression methods, which are difficult to deal with large-scale, high-dimensional data, and difficult to capture the nonlinear characteristics of price fluctuations. Therefore, an in-depth learning based forecasting method for commodity retail prices in international trade is proposed. The preparation of sample data is completed by acquiring and preprocessing the data of influencing factors of retail prices of international trade commodities, and a CNN-GRU model is constructed by combining the convolutional neural network and gating cycle unit in the in-depth learning, in which the sample data is input to realize the prediction of future retail prices of international trade commodities. The experimental results show that when the design method is used to predict the retail price of international trade commodities, it can accurately and stably predict the retail price of commodities, and achieve the expected effect.