Dynamic Pricing Models for Electronic Business

First, it is demonstrated that dynamic pricing may be used in different forms, dependent on the application at hand. For Instance, it may refer to the fact that different people may pay a different price for the same resource. But it could also be that dynamic pricing may refer to the fact that the per unit price may be different, based on the total quantity of units purchased (e.g., in quantity discounts). The models that have been used for addressing dynamic pricing, usually in the form of an optimisation problem, fall into one of the following categories. 

• Inventory-based models. These are models where the prices are determined by inventory levels and customer service levels. 

• Auction-based models. Auctions form a natural model for dynamic pricing, which take into account supply-demand characteristics of the market.  

• Other game-theoretic models. Scenarios with multiple providers essentially induce a game  where they compete for customers. Hence one can derive the prices by studying the equilibria of such games. 

• Data-driven models. These models use statistical techniques to compute optimal Dynamic prices.  

• Machine-learning models. This can be another natural class of models, where the service providers may opt to learn the customers’ behavioral patterns and deduce then optimal prices to maximise revenue. 

Finally, in the presence of stochastic demands, authors exhibit how one can employ a reinforcement learning based approach for dynamic price determination. In particular, it is shown that in a market with a single seller, the seller can use RL to modulate the prices dynamically so as to maximise the average profit or related performance metrics.