One Master Data Identification Method Based on Stacking Integration

In the existing master data identification process, there are several problems, such as slow demand response and lack of targets, after the construction of information systems. One master data identification method based on stacking integration is proposed in this paper. Considering the characteristics of master data and the available elements in the conceptual design stage, seven master data identifiable metrics are selected. Master data recognition result belongs to either the positive class or the negative one, which can be converted into so-called Binary Classification problem. In order to identify the characteristics of different attributes within the indicator, the first step is to learn with specific learners, the second step is to stack and integrate the results of each learner, and the last and third step is to obtain the results of identification. The experiments use both the self-built dataset and the Titanic dataset. The results show that this method can reach the expectations. The feasibility and validity of the method are verified as well.