Code reviewer recommendation system and method based on random forest classifier

Abstract

The invention discloses a code reviewer recommendation system and method based on a random forest classifier, and the system comprises an input module, a calculation module, a model training module, and a recommendation result output module, and a code reviewer recommendation method based on the random forest classifier is provided, and the method comprises the steps: inputting a historical code review record of a project; according to historical code review records, mining personnel activeness, code changes and file weight features; taking a reviewer of a historical project review record as aclassification label, taking the calculated feature vector as a data set, and inputting a training set into the random forest model to train a classification model; and extracting features from the to-be-evaluated code change, inputting the features into the classifier, and outputting N categories with the highest probability as recommended evaluators. According to the invention, in a large-scaleproject, an appropriate reviewer is recommended for the change of the to-be-reviewed code according to the historical review record, a reference basis is provided for the selection of the reviewer, and the communication cost is saved.

Description

technical field [0001] The invention relates to the technical field of software development, in particular to a code reviewer recommendation system and method based on a random forest classifier. Background technique [0002] As an important means of ensuring code quality, code review has become increasingly prominent in most software companies. However, the process of selecting reviewers usually takes a certain amount of communication time. How to find the right reviewer in time has become a code review An important problem in practice, sometimes the selection of reviewers is not reasonable enough, which may cause many problems for subsequent delivery. Appropriate code reviewers need to have certain relevant knowledge reserves and familiarity with the submitted code. At the same time, large-scale projects usually involve a large number of developers. How to select suitable reviewers from a large number of candidates has become a problem in software development practice. a ...

AI Technical Summary

Problems solved by technology

Rule-based methods can usually only dig out reviewer selection rules from a relatively single point of view for recommendation, and the recommendation effect is usually not ideal; while machine learning-based recommendation methods can comprehensively analyze reviewer selection strategies from multiple perspectives to improve the recommendation effect , but the currently commonly used reviewer recommendation rules are difficult to integrate into machine learning models

At present, some research attempts to integrate the most commonly used recommendation rules based on file path similarity as a feature into the machine learning model, but when converting the rules into features, there is a problem that the feature dimension is too large, which greatly increases the computational complexity.

Images