Revenue management is a process of maximizing revenue by optimizing product availability and price. This dissertation proposes a method with which capacity-constrained industries maximize their revenue by adopting dynamic pricing and learning multi-dimensional customer responses in revenue management. The first chapter of this dissertation deals with dynamic pricing which prices are constantly adjusted based on algorithms that take supply and demand, and other external factors into account. In particular, it considers a situation when a seller, facing uncertain demands, sells a single product in a finite horizon and actively adopts dynamic pricing and quantity discount schemes. Then, it analyzes the optimal strategy for `Buy One Get One Free’ and `50%’ off, both of which retailers often take. The rest of the dissertation focuses on the method for learning customer responses. Recently, data collection has become easier with the development of digital devices and machine learning that recognizes hidden patterns using data has rapidly emerged via deep learning. This has enabled us to learn multi-dimensional customer responses to the firm’s various promotional strategies. Using various modes of data including images, the second chapter proposes a deep learning method that forecasts customer demand. Chapter three presents a method for estimating customer trends in social network services with computer vision technology. Furthermore, using transaction data of individual customers, the final chapter suggests a new method of predicting the purchase probability of individual customer towards various types of products that the firm offers.