Dynamic pricing plays a vital role in modern e-commerce by enabling adaptive price adjustments that
maximize revenue in highly competitive markets. However, traditional machine learning techniques
often fail to capture the sequential and nonlinear dynamics of historical price data. To overcome this
challenge, we propose a hybrid deep learning model that integrates the feature extraction capabilities of
XGBoost with the temporal sequence modelling strength of Long Short-Term Memory (LSTM) networks.
By combining structured feature learning with time-dependent behavioural patterns, the model enhances
the accuracy of price prediction, demand forecasting, and elasticity estimation. Experimental results on
a synthetically generated e-commerce dataset show that the hybrid framework outperforms conventional
models, achieving a significant reduction in RMSE and improved R² scores. These findings highlight
the potential of hybrid deep learning approaches as a robust and scalable solution for implementing
intelligent, data-driven pricing strategies in both mobile commerce and traditional retail environments.
