- Developed Recommendation Algorithm that Allows Merchant with Insufficient Purchasing Information to Use Without Using User's Personal Information
- Follow-up Study of Real-Time, Natural Language Based Recommendation Algorithm that Optimizes Deep Learning and Reinforcement Learning Structure
- Contribution to the Improvement of Social Value on Artificial Intelligence
Harex InfoTech (Kyung-Yang Park, CEO), UBpay (User Centric Mobile Payment Shared Platform) Service Provider, announced the results of the user centric artificial intelligence (UCAI) research that has been conducted with professor Kyoung-Jun Lee, business school & big data application of Kyung Hee University, at the 2020 winter conference of The Korea Society of Management Information Systems in Seoul on December 17, 2020.
The research team defined the meaning of user centric artificial intelligence (UCAI) and proposed its structure that enables the development of an artificial intelligence system and the provision of appropriate services while preserving user privacy. The research team also developed an Extrapolative Collaborative Filtering (ECF) Algorithm that recommends products to customers by utilizing the user’s purchasing information including the user’s data from other merchants to complement and enhance the insufficient data from a single merchant perspective without utilizing the user’s personal information.
The research found that, without using personal information, the user centric artificial intelligence service can provide relevant product recommendation services to a user using only purchasing information such as purchased items, place of purchase and date while preserving the individual user’s privacy (Privacy-Preserving) while enabling secure collaboration between business entities without sharing the user’s data between those entities (Secure Collaboration).
Validating this algorithm with data held by Harex Infotech, the research team has developed the algorithm based on the vector (V-ECF) that does not treat products as standard categorized matrices. The team also developed a recommendation algorithm (N-ECF) that uses product names, store names, and purchase records in natural language, with excellent results. This means that it can make recommendations in an environment that includes different languages.
Professor Kyoung-Jun Lee said, " The N-ECF algorithm implemented with natural language utilization maximizes Multi-Merchant's Scalability, a characteristic of a payment service provider that covers multiple client stores so that Harex InfoTech Inc. expects to be able to provide a service with a single algorithm in any country regardless of the languages used.
Kyung-Yang Park, CEO of Harex InfoTech Inc., explained, "This research will be the starting point to further enhance the social value of artificial intelligence, which creates new transactions with differentiated Hyper-personalized On-Demand Service by analyzing the customers' daily transaction data with user centric artificial intelligence." He also said, "We plan to conduct follow-up research to include upgrading the real-time recommendation algorithm that optimizes deep learning and reinforcement learning structures so we may offer the AI service to the Korea Military Pay App scheduled to start in March 2021 and apply it abroad in places like Vietnam, Indonesia, Philippines, Myanmar, Saudi Arabia, Ethiopia, and Ghana, etc.”
Kyung-Yang Park, President and Chief Vision Officer (CVO) of Harex InfoTech, presented the concept of user centric artificial intelligence as a model for next-generation artificial intelligence that protects users' privacy and creates new transactions for the first time in the industry in 2016 at Seoul National University. Professor Kyoung-Jun Lee, the research head, published a paper designed for the user centric payment business model at the 8th International Conference on Electronic Commerce 2006 in Canada and won the Innovative Artificial Intelligence Application Awards three times from the Association for the Advancement of Artificial Intelligence (AAAI).
