Optimization of markets with AI


The circular economy that Mercari is aiming to create consists of a variety of complex markets. Our goal is to achieve safe and secure markets that we also stimulate as a participant. Both market forecasting and accurate decision-making are essential for us to achieve this goal. This project aims to realize safer, more secure, and more efficient markets by making such decisions at a high level. We plan to do this through AI research and development that leverages stochastic analysis and machine learning.


Our vision for the circular economy requires not only the simple exchange of goods, but also the formation of markets based on the cycle of production, distribution, and reuse. Furthermore, markets are now also diversifying to include cross-border e-commerce and digital asset markets such as crypto assets. Relying only on human processing capabilities limits our ability to accurately understand the conditions of those markets, as well as our ability to design and maintain safe, secure, and efficient markets. So, the use of computers is required. We expect that developing machine learning technology, particularly deep learning, will be useful in predicting the various events that are likely to occur in these diversifying markets. However, while computers are good at making predictions, they fall short at decision making. While deep reinforcement learning has proven successful at playing board games such as Go, it is still not ready to take on more advanced fields like market design and decision-making on the markets. This project aims to solve these problems by introducing concepts like stochastic analysis and economics in addition to machine learning techniques, and by leveraging data from the secondary distribution market, which is uniquely available to Mercari.


Optimal liquidation strategy for cryptocurrency marketplaces using stochastic control

Kenji KUBO, Kei NAKAGAWA, Daiki MIZUKAMI, Dipesh ACHARYA, Optimal Liquidation Strategy for Cryptocurrency Markets Using Stochastic Control, JSAI Technical Report, Type 2 SIG, 2022, Volume 2022, Issue FIN-029, Pages 23-27
Publication year: 2022