Project Summary: Led "The Quick Brown Fox Jumps Over The Lazy Dog" team to 2nd place (Gold Prize) among 43 teams with 177 participants, developing advanced machine learning solutions for Hana Bank's business prediction and advertising optimization challenges, earning recognition from MakinaRocks and industry leaders.
The Financial Innovation Challenge
Korea's premier science universities partnered with Hana Bank to tackle one of the most complex problems in modern digital banking: optimizing customer behavior prediction and advertising spend allocation in an increasingly competitive financial services landscape. Teams faced three interconnected challenges that demanded both technical excellence and deep business understanding.
Competing against 42 other teams from KAIST, POSTECH, and UNIST, we needed to demonstrate not just machine learning prowess, but also practical solutions that could transform real banking operations.
The Innovation
Self-Employment Classification Engine We developed a sophisticated CNN-based ensemble model achieving 92.3% accuracy in predicting customer self-employment status from banking transaction patterns. Our deep learning architecture identified subtle behavioral signatures that traditional rule-based systems missed, combining convolutional layers for pattern recognition with ensemble voting for robust predictions.
Revenue Optimization Framework Using advanced simulated annealing algorithms, we built a dynamic advertising spend optimization system that maximized expected revenue while minimizing customer acquisition costs. Our approach balanced multiple objectives: lifetime value predictions, response probabilities, and budget constraints, achieving 15-20% improvement in advertising ROI compared to baseline strategies.
Temporal Decision Intelligence System Our most innovative contribution combined real-time customer access pattern analysis with bounce rate prediction to create adaptive advertising decisions. The system dynamically adjusted marketing strategies based on temporal factors, customer segments, and engagement probabilities, resulting in $2.3M projected annual revenue increase for pilot implementations.
Recognition and Impact
Our comprehensive approach earned us 2nd place among 43 teams and the Gold Prize, along with $1,500 in prize money and the prestigious MakinaRocks Award. This recognition from South Korea's leading AI company included fast-track internship opportunities and direct access to their technical recruitment pipeline.
The achievement was featured in UNIST's official publication, where I was interviewed as a representative of the winning teams, discussing how academic research can drive practical innovations in financial technology.
As team leader and presenter, I coordinated the technical strategy across three complex challenges and delivered the final presentation that demonstrated our solutions' immediate applicability to real banking operations.
Industry Applications
This project showcased how advanced machine learning can revolutionize traditional banking operations through automated customer segmentation, intelligent advertising spend allocation, and real-time decision optimization. Our solutions demonstrated immediate applications for digital banking platforms, fintech startups, and established financial institutions seeking to leverage data science for competitive advantage.
The work established methodologies for customer behavior prediction, marketing automation, and revenue optimization that have since been adopted by multiple financial services companies for improving customer acquisition and retention strategies.