The invention discloses a commodity purchase quantity prediction method and device based on deep learning. The method comprises the steps of acquiring historical sale data of target commodities; according to the historical sale data, performing training to generate a neural network model; predicting a commodity sale trend of the target commodities in a future preset time period through the trainedneural network model to obtain a prediction result; and by utilizing the prediction result, performing calculation to obtain a purchase quantity of the target commodities. According to the method andthe device, on the basis of the prediction of the future commodity sale situation, the purposes of reducing commodity overstocking and improving operation efficiency of an e-commerce platform are achieved.