Graph database for transaction

• Designed and implemented graph database to speed up near real time BI query and to support multi-degree data integration

• Developed, monitored and maintained scalable cross-platform ETL data pipeline scheduled by Airflow.

• Built dashboard for database quality check and near real time integration level monitoring and analysis

• Design graph algorithm query to detect fraud relationship between fraud merchants and fake users for anti-fraud purpose

Comment analysis and fraud detection

• Designed analysis tool based on NLP for transaction comment to support fraud detection and context management

• Implemented machine learning model and NLP algorithm to build classifier for comments sentiment analysis, fraud detection and topic identification.

• Built ETL data pipeline scheduled by Airflow for automatic data processing and results storage

• Achieved 88% accuracy on sentiment analysis and 100% accuracy on top 5% merchant level fraud detection