¹Department of Computer Science, University of Uyo, Uyo, Akwa Ibom State, Nigeria.
²Department of Computer Science, University of Benin, Benin City, Nigeria.

---

Abstract—Adaptability is an important software quality characteristic and a major non-functional requirement in software and should therefore be given adequate attention during software quality measurement and predictions especially now that the environment in which software products operate becomes highly unpredictable due to rapid changes in hardware platform as well as changes in the operating system requirements. In this work the results from the analysis of object-oriented software source code using our previously developed software analyser to measure the values of some internal properties using object-oriented software metrics based on formulation of decision rules in conjunction with binary logic combination of the possible internal software properties which aided in the prediction of adaptability level of a given software, is used as the dataset for ordinal logistic regression analysis. This analysis was used to formulate the adaptability model based on proportional odds assumption. The results showed that software with low coupling, inheritance and complexity were more likely to be adaptable than those with high values. Conversely, software with low cohesion were less likely to be adaptable than those with high cohesion. High cohesion was associated with adaptability since its odds ratio (low/high) was 7.97 (>1) while low coupling, inheritance and complexity were associated with adaptability since their odds ratios (low/high) were 0.15, 0.22 and 0.05 (<1) respectively. The resulted model fitted the data well and the estimated cumulative odds were the same across all the ordinal categories, thus the proportional odds assumption held.

Index Terms—Metrics, model, ordinal logistic regression, proportional odds, software adaptability.

[PDF]

Cite: Udo, E. N., Akwukwuma, V. V. N., "Software Adaptability Metrics Model Using Ordinary Logistic Regression," Journal of Software vol. 14, no. 3, pp. 116-128, 2019.