Young Shin Han, an AI big data major at Kookmin University's Graduate School of Business, and Joo Hyun Kim, a business administration major at the Graduate School of Management (advisor Moon Hyun-real), recently won the Best Paper Award at the 2024 Korean Academy of Management Science Fall Conference held at Korea Aerospace University.

Youngshin Han presented a paper titled “A Method for Generating a Summary of Participant Reactions in Real-time Personal Broadcasting Using RAG”. Since real-time personal broadcasts are broadcast without a set time, it is difficult for viewers to participate in the broadcast according to the broadcast time. Therefore, it is necessary to summarize past broadcasts, but previous research has focused on extracting specific highlight scenes, which limits the ability to summarize broadcasts that reflect the entire context of the broadcast. To solve this problem, Youngshin Han proposed a methodology for summarizing broadcast contents using Retrieval-Augmented Generation (RAG) considering the characteristics of real-time personal broadcasting. In particular, unlike traditional mass media, real-time personal broadcasting is a two-way communication between participants and streamers through chat, so a new methodology is proposed to summarize the entire broadcast chat to summarize the overall flow and context of the broadcast. The key achievement of the study is that summarization using RAGs improves the quality of summaries compared to summaries using only LLM (Large Laguage Model). In addition, the proposed summarization method performed well in terms of speed and efficiency. Youngshin Han's research is expected to make practical contributions to the media content industry in the future by providing an efficient way to utilize data generated from real-time personal broadcasts.

Joohyun Kim presented a paper on “Mobile ad targeting method based on Uplift model”. Recently, personalized advertising using log data has been attracting attention, but targeting based on predictive models has the limitation that it is likely to include customers who are not targeted by ad promotions, and A/B testing is time-consuming and expensive. To solve this problem, Joohyun Kim proposed a methodology that applies the Uplift model to measure behavioral changes caused by ad exposure based on data and target only those consumers who are likely to convert. In her research, she implemented MOA (Modified Outcome Approach) and MetaLearner's T-learner and S-learner with RandomForest and XGBoost algorithms to build six models to measure the net effect of ads by customer group. The key finding of the study is that Uplift modeling can identify the optimal customer groups that can achieve high ROI. In addition, by analyzing the characteristics of the customer groups to be targeted in-depth, Uplift modeling further enhanced the efficiency of data-driven targeting. In addition to contributing to the academic development of marketing and data science, Joohyun Kim's research is expected to bring practical results in maximizing marketing effectiveness in various industries through targeting techniques using the Uplift model.