Trans-company software defect prediction method based on transfer learning and defect quantity information

Abstract

The invention discloses a trans-company software defect prediction method based on transfer learning and defect quantity information. The method comprises five steps of marking the number of defects of each trans-project case, namely training data; extracting the measurement attribute in the case; preprocessing data; constructing a Bayes defect prediction model based on a weighted trans-project case set; and predicting whether a current case has defects according to the Bayes defect prediction model. The invention improves a conventional trans-project defect data weight calculation method, and provides the trans-company software defect prediction method based on transfer learning and defect quantity information. The trans-company software defect prediction method takes account of the addition effect of the defect quantity information on the basis of the weight calculation according to the transfer learning, and avoids influences of imbalance problems on prediction results so as to improve the accuracy of the trans-project defect prediction.

Description

technical field [0001] The invention belongs to the technical field of cross-company software defect prediction, in particular to a cross-company software defect prediction method based on transfer learning and defect quantity information. Background technique [0002] (1) Software project defect prediction technology [0003] As an important part of the information industry, the software industry plays an important role in promoting the adjustment of industrial structure, the transformation of economic development mode, the integration of industrialization and information technology, and the maintenance of national security. However, with the continuous expansion of software application fields and the continuous improvement of design complexity, the quality of software projects is becoming more and more difficult to control. In the context of this demand, efficient software defect prediction technology has attracted more and more attention. [0004] Software defect predic...

AI Technical Summary

Problems solved by technology

[0018] Aiming at the problem that the weight ranking cannot be determined when the class imbalance and the distance between the data are equal in the existing cross-item defect prediction method based on transfer learning, that is, when the two instances and the data to be predicted have the same similarity , which instance should be given a higher weight, or, in the class imbalance problem, the accumulation of a large number of non-defective instance weights will cause the weight of defective instances to have little effect on the result, how to eliminate this adverse effect problem, the present invention provides a cross-company software defect prediction method based on transfer learning and defect quantity information

Images