Software Quality Prediction Based on Generalized Linear Models  
Author Shinji Inoue

 
Co-Author(s) Shigeru Yamada

 
Abstract Statistical analysis approaches are often conducted for software quality prediction based on software process data collected from an actual software development project. As one of the main approaches is to use multiple regression analysis where the objective variable represents the number of faults detected in the system testing and the software process factors are treated as the explanatory variables. However, this approach is unsuitable because the objective variable is permitted to take the negative value in the multiple regression approach. We propose another statistical approaches for overcoming the problem by applying the generalized linear modeling approach to the software quality prediction. Further we check the prediction and fitting performance of our approaches with existing approach by using actual software process data.

 
Keywords Software quality/reliability, Software process data, Statistical analysis, Generalized linear models, Model comparisons
   
    Article #:  2194
 
Proceedings of the 21st ISSAT International Conference on Reliability and Quality in Design
August 6-8, 2015 - Philadelphia, Pennsylvia, U.S.A.