Accurate demand forecasting in online retail is a fundamental challenge in supply chain management and inventory planning, as demand patterns are typically influenced by complex temporal relationships, cross-product network interactions, and item-specific semantic features. Many traditional time-series approaches and even certain deep learning models struggle to simultaneously capture these dynamic, non-linear dependencies. To address this challenge, this study introduces a novel forecasting framework based on a Transformer architecture, capable of integratively modeling long-lag temporal dependencies, cross-product network effects, and item semantic information. This empirical-computational research evaluates the proposed model using two real-world online retail datasets. Following data preprocessing and the removal of incomplete records, the data were partitioned into training and testing sets. The proposed model was subsequently trained utilizing a multi-head attention mechanism within the Transformer framework, incorporating semantic representations extracted from product descriptions. The results demonstrate that the proposed Transformer model achieves a Mean Absolute Error (MAE) of 2.68 and a Root Mean Square Error (RMSE) of 3.85, outperforming both the Autoregressive Integrated Moving Average (ARIMA) model (MAE: 4.37, RMSE: 5.96) and the Gradient Boosting algorithm (MAE: 3.91, RMSE: 5.12). This improvement indicates that the proposed framework effectively identifies complex demand patterns and cross-product relationships.