In the cost management of supply chain, inventory management is particularly important. In order to help suppliers better manage inventory and reduce inventory turnover cost, this paper will study the commodity sales forecasting model based on random forest regression algorithm and genetic algorithm in the field of machine learning, so as to dynamically optimize and guide inventory management. First, variance selection method is used for the initial feature selection. Then the data set is trained and predicted using the random forest regression algorithm, and the feature importance is sorted, and the second feature selection is carried out. Then genetic algorithm is used to optimize the hyperparameters of random forest regression algorithm to get the best combination of hyperparameters. Finally, the optimal combination of hyperparameters is used to construct a model for the selected important features, and the sales volume of the product is forecasted. By comparing the evaluation indexes of the model, it is found that using the optimized parameters and important feature data for training, the prediction effect is better.