| Abstract: | Pricing plays a central rule to a company’s profitability, and therefore has been extensively studied in the literature of economics. When designing a pricing mechanism/ model, an important principle to consider is “price discrimination”, which refers to selling the same resources with different prices according to different values of buyers. To meet the “price discrimination” principle, especially when the number of buyers is large, computational methods, which act in a more accurate and principled way, are usually needed to determine the optimal allocation of sellers’ resources (whom to sell to) and the optimal payment of buyers (what to charge). Nowadays, in the Internet era in which quite a lot of buy and sell processes are conducted through Internet, the design of computational pricing models faces both new challenges and opportunities, considering that (i) nearly realtime interactions between people enable the buyers to reveal their needs and enable the sellers to expose their information in a more expressive manner, (ii) the large-scale interaction data require powerful methods for more efficient processing and enable the sellers to model different buyers in a more precise manner. In this paper, we review recent advances on the analysis and design of computational pricing models for representative Internet industries, e.g., online advertising and cloud computing. In particular, we introduce how computational approaches can be used to analyze buyer’s behaviors (i.e., equilibrium analysis), improve resource utilization (i.e., social welfare analysis), and boost seller’s profit (i.e., revenue analysis). We also discuss how machine learning techniques can be used to better understand buyer’s behaviors and design more effective pricing mechanisms, given the availability of large scale data. Moreover, we make discussions on future research directions on computational pricing, which hopefully can inspire more researchers to contribute to this important domain. |
