Financial Big Data Reconciliation Method

For data errors in distributed financial system caused by multi-system interaction, asynchronous processing and system bug, this paper proposes offline and quasi real-time data reconciliation methods based on the combination of Alibaba big data processing platform and accounting theory. In offline data reconciliation, Full data reconciliation and hour level incremental data reconciliation are introduced. And in quasi real-time data reconciliation, single system and distributed multi-system reconciliation models are introduced. These data reconciliation methods are then verified against 7 million pieces of daily data of the distributed loan system in a financial company. Results show that these methods can complete the financial big data processing, discover the data quality problems timely, and minimize the financial system capital loss.