"Financial services
Building a Big Data platform for fraud detection
Starting point
- There are no self-learning fraud detection mechanisms that examine transaction data for fraud patterns in real time
- Big Data, and specifically Hadoop, are considered the foundation for a modern database architecture. The client wants to integrate the Hadoop ecosystem effectively into its existing enterprise architecture
- The basis for implementation is the successful pilot project of the use case 'Fraud Detection', which was implemented with regard to the existing security and operational infrastructure
Procedure
- Define interfaces with Hadoop
- Evaluate and implement reporting with IBM Cognos on the Hadoop platform
- Use of Hadoop streaming components for 'Fraud Detection'
- Development and handover of the IT operations concept, including hardware specifications
- Evaluation and implementation of the high availability and backup concept
- Integration of the platform into the existing security infrastructure
Features/Project outcome
- Real-time evaluation of financial transactions based on a statistical model; implemented with Apache Storm and Kafka
- Model development with Apache Spark and R
- Seamless modification of the evaluation model
- Result storage in Apache Hive and HBase for further analysis and evaluation
Customer benefits
- Reduction of fraud cases in financial transactions
- Evaluation of the transaction during the payment process
- Faster response to fraud cases
- Greater development and analysis capabilities for improving the evaluation model
- Relief of the existing enterprise architecture and RDBMS
- Cost reduction through more affordable storage in Hadoop
Rainer Unsöld | Member of Executive Board