Artificial Intelligence and Dynamic Pricing Strategies: Enhancing Competitiveness in E-Commerce

Dynamic pricing is vital for e-commerce companies to adapt to the dynamic pricing techniques if they have to maximize their revenues without having too much stock. To efficiently respond to changes in demand, accurate demand forecasting is central to the process of making changes in pricing. The main problem with conventional approaches to price determination is that usually there is no way the price can be changed in response to changes in the market. New developments in Fourier Analysis Networks (FANs) and Reinforcement Learning (RL) present the potential for improving demand forecasts and price optimization. The objective of this work is to implement FANs for demand forecasting in parallel to an RL agent for dynamic pricing and fine-tune the RL agent via Hyperband as the hyperparameter tuning method to create an efficient and adaptive pricing model. Thus, the integration of time series demand forecast with dynamic pricing mechanism allows for daily business to change their prices according to the forecasted demands, market status, and competitive prices. The hybrid system takes the demand forecast from the FANs to facilitate the RL agent's decision to set appropriate prices. RL agent acts with the environment, changing prices based on the forecasts of demand and stock, as well as the prices of competitors. Hyperband optimization, therefore, refines the hyperparameters of the agent to increase efficiency. The findings show how the implementation of the proposed hybrid system exceeds conventional pricing models. The demand forecast accuracy of the model stood at 95.65 % to achieve better revenues uplifts. The presented framework provides the e-commerce companies with the opportunity for better adjustment and implementation of different pricing strategies to increase profitability and minimize the time spent on identification of different conditions on the market.