Self-acceptance technology debt detection and classification method based on multi-method ensemble learning

Abstract

The invention relates to a self-acceptance technology detection and classification method based on multi-method ensemble learning. The method comprises the following five steps: preprocessing featurewords; selecting the first k most useful features to train a classifier; training corresponding sub-classifiers by using a naive Bayes polynomial method and a linear Logistic regression method; and performing integrated prediction on the prediction result through a sub-classifier voting rule to obtain accuracy, a recall rate, comprehensive accuracy and a recall rate, and finally calculating an F1value as a subsequent evaluation standard. And finally, the features which frequently appear in the experiment process and have high information gain values are clustered through a clustering method,so that the detected technical debts are classified.

Description

technical field [0001] The invention belongs to the technical field of software development, and in particular relates to a method for classifying and detecting self-admitted debts based on multi-method integrated learning. Background technique [0002] The document "Huang Q, Shihab E, Xia X, et al. Identifying self-admitted technical debt in open source projects using text mining [J]. Empirical Software Engineering, 2017." discloses an automatic detection of self-admitted technical debt using an integrated classifier Methods. The method uses source code annotations from different software projects to analyze the annotations to be detected. This method firstly preprocesses the source files, and uses feature selection to filter features, using Naive Bayesian polynomial ( BayesMultinomial) trains each classifier, and finally the integrated classifier composed of multiple classifiers predicts according to the voting rules to determine whether the statement has self-admitted ...

AI Technical Summary

Problems solved by technology

But so far some methods rely heavily on manual detection, and there are many advanced methods that use a single natural language detection method to automatically identify SATD. However, the manual detection method has low efficiency and obvious disadvantages, and the single natural language detection method classifier has low performance. , poor flexibility, although the previous detection also had good results, but the diversity of SATD in the project and the characteristics of semantic changes have brought great challenges to the accurate detection of SATD

Images