An inference structure for process feedback: technique and implementation

Abstract

This paper presents an automatic technique for making simple inferences about the stages in a software production process, discusses implementation of the technique, and validates the technique using defect data from several software development projects. The technique represents an approach to automate process feedback that may be based on either experience and common sense or historical data. Specifically, we present (1) A software defect classification scheme that relates defects to process stages. As an example of such a scheme, we describe 'orthogonal defect classification', which associates types of defects with stages, such as high-level design, low-level design, code, and system test. These associations are summarized in a 'principal association' table, which serves as a model for the production process. (2) Rules that provide simple in-process feedback about a process stage based on the defects associated with that stage. We discuss how inferences can be made, with and without historical information for the process, and illustrate application of the inference technique to the software production process. (3) An implementation of the rules using a statistical programming language. Finally, we consider the technique in the software production context as a particular case of a general technique that makes use of both human judgment and historical data, and discuss extensions. © 1994 Chapman & Hall.