arXiv subjects

Andrea Capiluppi

Publications and source records attributed to Andrea Capiluppi.

20 records · Page 2Linked to original sources

arXiv2021

Using Structural and Semantic Information to Identify Software Components

Component Based Software Engineering (CBSE) seeks to promote the reuse of software by using existing software modules into the development process. However, the availability of such a reusable component is not immediate and is costly and time consuming. As an alternative, the extraction from pre-existing OO software can be considered. In this work, we evaluate two community detection algorithms for the task of software components identification. Considering `components' as `communities', the aim is to evaluate how independent, yet cohesive, the components are when extracted by structurally informed algorithms. We analyze 412 Java systems and evaluate the cohesion of the extracted communities using four document representation techniques. The evaluation aims to find which algorithm extracts the most semantically cohesive, yet separated communities. The results show a good performance in both algorithms, however, each has its own strengths. Leiden extracts less cohesive, but better separated, and better clustered components that depend more on similar ones. Infomap, on the other side, creates more cohesive, slightly overlapping clusters that are less likely to depend on other semantically similar components.

cs.SE↗

arXiv2019

The Prevalence of Errors in Machine Learning Experiments

Context: Conducting experiments is central to research machine learning research to benchmark, evaluate and compare learning algorithms. Consequently it is important we conduct reliable, trustworthy experiments. Objective: We investigate the incidence of errors in a sample of machine learning experiments in the domain of software defect prediction. Our focus is simple arithmetical and statistical errors. Method: We analyse 49 papers describing 2456 individual experimental results from a previously undertaken systematic review comparing supervised and unsupervised defect prediction classifiers. We extract the confusion matrices and test for relevant constraints, e.g., the marginal probabilities must sum to one. We also check for multiple statistical significance testing errors. Results: We find that a total of 22 out of 49 papers contain demonstrable errors. Of these 7 were statistical and 16 related to confusion matrix inconsistency (one paper contained both classes of error). Conclusions: Whilst some errors may be of a relatively trivial nature, e.g., transcription errors their presence does not engender confidence. We strongly urge researchers to follow open science principles so errors can be more easily be detected and corrected, thus as a community reduce this worryingly high error rate with our computational experiments.

cs.LG↗