Kakao Pay's User Profiling System, PayProfile

PayProfile is a user profiling system that collects, processes, and classifies traces of Kakao Pay users' behavior to provide personalized services. It aims to analyze users' behavioral characteristics and predict future behavior through vast amounts of data.

Data composition goes through the following 4-stage process:

• Timeline: Extracts 695 Features from 9 domains and loads the data.
• Profiling: Standardizes the collected figures and scores them so that relative tendencies compared to other users can be identified.
• Modeling: Predicts users' unrecognized needs, such as life events (marriage, childcare, etc.) and interest in financial management.
• Segment: Creates user segments by combining attributes and tags, and links them with the CXM Customer Analytics Platform for utilization.

The core technology underpinning the system is the merchant classification model. It extracts consumption category, brand, and location information through deep learning, and in particular, location information secured scalability by utilizing an entropy index.

In addition, sophisticated modeling is performed through network analysis that identifies relationships between users, BERT-based natural language processing (NLP), and user clustering technology utilizing GNN (Graph Neural Network).