| 1. |
- Lavesson, Niklas, et al.
(författare)
-
A method for evaluation of learning components
- 2014
-
Ingår i: Automated Software Engineering. - : Springer. - 0928-8910 .- 1573-7535. ; 21:1, s. 41-63
-
Tidskriftsartikel (refereegranskat)abstract
- Today, it is common to include machine learning components in software products. These components offer specific functionalities such as image recognition, time series analysis, and forecasting but may not satisfy the non-functional constraints of the software products. It is difficult to identify suitable learning algorithms for a particular task and software product because the non-functional requirements of the product affect algorithm suitability. A particular suitability evaluation may thus require the assessment of multiple criteria to analyse trade-offs between functional and non-functional requirements. For this purpose, we present a method for APPlication-Oriented Validation and Evaluation (APPrOVE). This method comprises four sequential steps that address the stated evaluation problem. The method provides a common ground for different stakeholders and enables a multi-expert and multi-criteria evaluation of machine learning algorithms prior to inclusion in software products. Essentially, the problem addressed in this article concerns how to choose the appropriate machine learning component for a particular software product.
|
|
| 2. |
- Liparas, Dimitris, et al.
(författare)
-
Applying the Mahalanobis-Taguchi Strategy for Software Defect Diagnosis
- 2012
-
Ingår i: Automated Software Engineering. - : Springer Science and Business Media LLC. - 1573-7535 .- 0928-8910. ; 19:2, s. 141-165
-
Tidskriftsartikel (refereegranskat)abstract
- The Mahalanobis-Taguchi (MT) strategy combines mathematical and statistical concepts like Mahalanobis distance, Gram-Schmidt orthogonalization and experimental designs to support diagnosis and decision-making based on multivariate data. The primary purpose is to develop a scale to measure the degree of abnormality of cases, compared to “normal” or “healthy” cases, i.e. a continuous scale from a set of binary classified cases. An optimal subset of variables for measuring abnormality is then selected and rules for future diagnosis are defined based on them and the measurement scale. This maps well to problems in software defect prediction based on a multivariate set of software metrics and attributes. In this paper, the MT strategy combined with a cluster analysis technique for determining the most appropriate training set, is described and applied to well-known datasets in order to evaluate the fault-proneness of software modules. The measurement scale resulting from the MT strategy is evaluated using ROC curves and shows that it is a promising technique for software defect diagnosis. It compares favorably to previously evaluated methods on a number of publically available data sets. The special characteristic of the MT strategy that it quantifies the level of abnormality can also stimulate and inform discussions with engineers and managers in different defect prediction situations.
|
|
