Comprehensive Prediction Method of Multi-feature Software Defects Based on Unbalanced Noise Set

Abstract

The invention discloses a multi-feature software defect comprehensive prediction method based on an unbalanced noise set, comprising the following steps: constructing an initial data set including code features, development process features and network structure features. Initially undersample the dataset to reduce duplicates in the majority class. The k-nearest neighbor sample set is found for the metric element in the dataset by means of propensity score matching. Through the k-nearest neighbor sample threshold judgment, the noise reduction processing of the data set is realized. Synthesize the minority class in the data set and the minority class in the k-nearest neighbor sample set to eliminate the class imbalance problem in the data set. Adaptively build a variety of machine learning models and select the most suitable machine learning model to predict defects for new versions of software. The invention solves the class imbalance problem commonly existing in software defect prediction. Noise samples are removed through noise discriminant processing based on propensity score matching.

Description

Technical field [0001] The present invention relates to software defect prediction and software reliability technology, and in particular, to the comprehensive prediction method of multi-featured software defects based on unbalanced noise sets. Background technique [0002] In the 21st century, the status of software in social life has become increasing, everywhere, is not limited to engineering, scientific research, economy and other professional fields, and has entered thousands of households, becoming a pivotable tool in people's lives. With the increasing software system, the complexity is geometric, and the shortcomings of software are also increasing, and software defects have also affected all aspects of the people's lives. [0003] Software defect data is the basis for software reliability related research and application. For a long time, defective data used in software reliability is mainly from limited software test data and analysis data in use. Although this type of ...

AI Technical Summary

Problems solved by technology

[0009] The purpose of the present invention is to propose a comprehensive prediction method for multi-feature software defects based on an unbalanced noise set, which can solve the problems caused by incomplete feature measurement, high degree of unbalanced data sets, and wrong labels in the data set in the prior art. noise samples etc.

Images