Deep Learning and Optimization Approaches in E-commerce Systems for Sale Prediction Using Triple Pseudo-Siamese Network with Giant Trevally Optimizer: A Review

Abstract

The rapid growth of e-commerce platforms has resulted in the generation of vast amounts of transactional and behavioral data, enabling the use of predictive analytics for sales forecasting and recommendation systems. Accurate sales prediction is essential for effective inventory management, demand planning, and strategic decision-making in online retail. However, traditional statistical models often fail to capture the nonlinear relationships and complex patterns present in high-dimensional datasets. To overcome these limitations, deep learning approaches have been widely adopted, with Siamese and pseudo-Siamese neural networks gaining prominence due to their ability to learn similarity relationships across diverse feature spaces. The Triple Pseudo-Siamese Network further enhances this capability by integrating multiple feature streams, including customer behavior, product attributes, and temporal sales patterns, to improve prediction accuracy. Efficient training of such deep models requires robust optimization techniques, and metaheuristic algorithms like the Giant Trevally Optimizer (GTO) have shown strong potential by effectively exploring complex search spaces. By combining advanced neural architectures with optimization strategies, modern approaches significantly enhance forecasting performance. This review highlights recent developments, identifies research gaps, and outlines future directions for scalable e-commerce prediction systems.