A Relevant Patch Recommendation Method Based on Heterogeneous Data

Abstract

The invention discloses a method for recommending related patches based on heterogeneous data, which includes crawling multiple heterogeneous data for code review, cleaning the data, splicing the features of multiple heterogeneous data into patch feature vectors, and pairing patch pairs so as to correlate with predicted patches. The correlation is a positive sample, and the non-correlation with the predicted patch is a negative sample. After marking the positive and negative samples with a binary classification label, divide the training set and the verification set, and use the training set to train the three models of logistic regression, random forest and LightGBM respectively. , get the corresponding probability and prediction label, then calculate the corresponding accuracy rate according to the prediction label, and finally construct the prediction score according to the weighted sum of the fusion weight and the corresponding probability, and obtain the optimal prediction score. The present invention utilizes machine learning to conduct correlation evaluation on the data submitted to the code review system to obtain the optimal recommendation, improves the reliability and stability of the recommendation, and saves labor costs.

Description

technical field [0001] The invention relates to the field of code review, in particular to a related patch recommendation method based on heterogeneous data. Background technique [0002] Code review is an important basis for the smooth iteration of software engineering projects. It is composed of multiple and complex small tasks, including code specification revisions, code supplementary comments, etc. At present, it is widely used in the field of software engineering to update the code and manage the version by manual review, and the labor cost is high. [0003] At present, the software engineering industry generally uses git, Gerrit and other similar systems for code management and review. Every time the code is updated on these systems, we call it a code modification or a patch. These systems provide a reviewer, committer, for each patch submission to complete queries or comments. The basic process for patch review is: [0004] 1) The programmer, that is, the author ...

AI Technical Summary

Problems solved by technology

However, due to the large number of code iterations and too many code files in each project, it takes a lot of labor cost and time to manually find other submissions related to the current problem submission, or to manually mark whether the submissions are related in advance. cost

Images