In the big data platform warehouse system, both the integration layer and the foundation layer are built based on the ten major thematic models of the financial sector. These are further customized and developed in accordance with the specific characteristics of banking data, forming a dual-framework paradigm architecture of “logical model + physical model.”
For the logical data model, the main entity models are decomposed according to indicators such as data volume, usage frequency, and data update frequency. A dual-entity approach is adopted in logical model design to ensure both the standardization of the logical model and the timeliness required for large-scale banking data.
For the physical model, based on metrics such as data volume and batch processing duration, physical partitioning and model preprocessing are carried out. This enhances both the efficiency and timeliness of data processing.

In the common processing layer of the big data platform warehouse system, the bank’s fundamental indicators are uniformly processed and calculated across five dimensions: agreement, customer, channel, organization, and product, thereby providing standardized data services.
For data indicators with the same business meaning, the common processing layer ensures a unified and unique standard is applied during implementation. The processing specifications and indicator names are incorporated into the bank’s data standards framework and released across the organization in a consistent manner.

In the data mapping process of the big data platform warehouse system, data code values are aligned with the bank-wide published data standards, field names are standardized, and identical code values from different source systems are consolidated. This reduces data redundancy and improves both data usage efficiency and user experience.
Fields with poor data quality, high null rates, or ambiguous meanings are filtered out and flagged for feedback, thereby enhancing the quality of data available for downstream applications.

By adopting open-source frameworks and developing innovative visualization components, a data visualization development framework and standards tailored for the banking sector have been established. Within this framework, map elements from the original development environment have been enhanced to include visualized displays of customer migration trajectories. In addition, multiple models such as graph theory, time series, random forest, and text mining are leveraged to develop diverse data products for model R&D.

Develop data products using a variety of models such as graph theory, time series, random forest, and text mining.