User Portrait Technology Based on Stacking Mode

Abstract

Because of the differences in code styles and programming levels among developers, it is prone to code irregularities, poor readability, and security vulnerabilities. Although developers can see their problems in the test report, it is difficult to guarantee that they will not make mistakes on the same problem. In this article, we propose a way to build a programming level for developers. The proposed method explores the Stacking model for building a programming level for developers. First, cluster the problems that occurred during the development process, give weight to the category information, score the developer’s programming level, and divide the user groups into different categories. Then use the word vector to extract the features of the code defect, generate a feature matrix, and finally pass the feature matrix to the Stacking classifier to classify the defect information and update the developer’s programming level portrait. Experimental results show that it is effective in predicting defect code information. In addition, a comparative study with the state-of-the-art method shows that the method is superior to existing methods in terms of classification rate, recall, precision and F-measure.

Yirui Wu

Young Professor, CCF Senior Member

My research interests include Computer Vision, Artifical Intelligence, Multimedia Computing and Intelligent Water Conservancy.