부산대학교 도서관

E-commerce is the practice of purchasing and selling goods over the Internet. Customers will appreciate the simplicity of not having to make physical purchases using e-primary commerce. They will order the item online and deliver it to their home as soon as feasible. This research aimed to create deep learning algorithms capable of forecasting e-commerce sales. This project aims to develop and test a model for predicting online product sales over a wide range of online product types. It bases its decisions on the requirements for online product sales, the factors influencing online product sales in various industries, and the benefits of the deep learning algorithm. A continuous Stochastic Fractal Search (SFS) method for optimizing the parameters of a deep learning-modified neural network (DLMNN) is introduced in this paper. In e-commerce demand forecasting studies, a time series dataset is also analyzed. The DLMNN model’s performance improvements across multiple product categories are illustrated using a non-deep learning model as a baseline comparison. The experiment also shows that the unsupervised pretrained DLMNN model outperforms the competition in terms of sales predictions. For root mean square error, the proposed technique obtained RMSE, Mean, and Standard Deviation.