A Robustness Analysis of Imputation Method for Software Development Project Data: Missing Value Treatment for Software Quality Prediction
Takayuki Morita, Mitsuhiro Kimura · International Journal of Software Engineering and Its Applications · 2015
As our goal, we are interested in estimating the degree of software reliability based on software development project data. It is widely-known that several software development attributes which are measured can be used to evaluate and predict software reliability/quality via multi-variable analyses. In this article, we focus on the data treatment method which is needed prior to the software reliability assessment, since the software development data sets often include missing data. This paper discusses the method of data preparation against missing data and their effectiveness by using the Random Forest as a multi-variable analysis.