Using CBR and CART to predict maintainability of relational database-driven software applications

摘要: Background: Relational database-driven software applications have gained significant importance in modern development. Given that maintainability is an important quality attribute, predicting these applications' can provide various benefits to organizations, such as adopting a defensive design and more informed resource management. Aims: The aim of this paper present the results from employing two well-known prediction techniques estimate relational applications. Method: Case-based reasoning (CBR) classification regression trees (CART) were applied data gathered on 56 projects companies. concerned development and/or maintenance Unlike previous studies, all variables (28 independent 1 dependent) measured 5-point bi-polar scale. Results: Results showed CBR performed slightly better (at 76.8% correct predictions) terms accuracy when compared CART (67.8%). In addition, predictors identified documentation understandability Conclusions: show be used by companies formalize improve their process prediction. Future work involves gathering also other techniques.