Establishment of quality evaluation index system of machine learning model

• 1、School of Cyberspace Security, Beijing University Of Posts And Telecommunications, Beijing 100876

Abstract：Objective To explain the quality of machine learning model and construct a comprehensive and reasonable machine learning model quality evaluation index system. Methods Use literature research method to conclude the principles and methods of index system construction, summarize the machine learning model features, build the machine learning model quality model. Use content analysis to construct the initial evaluation index system. Use qualitative analysis to preliminarily filter the index. Use expert research to select the important index. Use the retested confidence test to test the credibility of the index. Use analytic hierarchy process to determine index weight. Result This paper proposes a hierarchical quality model for machine learning model, and proposes machine learning model quality evaluation index system which contains 6 first-level indexes, 7 second-level indexes and 5 third-level indexes, and build the handwritten digital recognition model quality evaluation index system. Conclusion In this paper, the index system is comprehensive and reasonable, and the number of index is suitable, and also has strong operability in specific scenarios.

Keywords： Fundamentals of computer science and technology Evaluation index system Machine learning model Quality model Analytic hierarchy