Data mining method with model early-warning updating mechanism

Abstract

The invention discloses a data mining method with a model early-warning updating mechanism, and relates to the field of CRISP-DM (cross-industry standard process for data mining). The method includes:1, obtaining a preliminary scheme of a service objective by commercial understanding, and defining a model failure determination principle according to the service objective; 2, carrying out data understanding and data preparation in sequence on the basis of the preliminary scheme to obtain a data set suitable for modeling analysis; 3, training multiple models on the basis of the data set and thefailure determination principle to complete establishment and optimization of the models; 4, carrying out model evaluation and preliminary deployment on the multiple established models, then judgingwhether the same meet an early-warning rule and need to be updated, if the same meet the early-warning rule, recalculating the models to complete updating, and then jumping to a step 5, and if the same do not need to be updated, directly jumping to the step 5; and 5, carrying out final deployment of the models to complete data mining. According to the method, problems of low precision and high costs brought by repeated execution processes caused by existing data mining processes due to smaller model numbers are solved, and effects of improving model precision and reducing costs are achieved.

Description

technical field [0001] The invention relates to the field of cross-industry data mining standard procedures, in particular to a data mining method with a model early warning update mechanism. Background technique [0002] CRISP-DM (cross-industry standard process for data mining) "Cross-industry standard process for data mining" is a data mining methodology jointly developed by NCR, OHRA, SPSS, Daimler-Benz and other global companies. Compared with other existing data mining methodologies , the CRISP-DM methodology is more superior, so it is widely used. [0003] The CRISP-DM methodology defines data mining practice as six standard stages, which are business understanding, data understanding, data preparation, model building, model evaluation, and model deployment. The following are brief introductions: [0004] 1. Business understanding: [0005] Business understanding is to clarify the business goals to be achieved and transform them into data mining topics; understand t...

AI Technical Summary

Problems solved by technology

[0017] The purpose of the present invention is: the present invention provides a data mining method with a model early warning update mechanism, which solves the low accuracy and high cost of the existing data mining process due to the small number of models and the lack of actual data inspection resulting in repeated execution of the process The problem

Images