Investigation of automatic prediction of software quality

The subjective nature of software code quality makes it a complex topic. Most software managers and companies rely on the subjective evaluations of experts to determine software code quality. Software companies can save time and money by utilizing a model that could accurately predict different code quality factors during and after the production of software. Previous research builds a model predicting the difference between bad and excellent software. This paper expands this to a larger range of bad, poor, fair, good, and excellent, and builds a model predicting these classes. This research investigates decision trees and ensemble learning from the machine learning tool Weka as primary classifier models predicting reusability, flexibility, understandability, functionality, extendibility, effectiveness, and total quality of software code.