Jerffeson Teixeira de Souza

Ten years of search based software engineering: a bibliometric analysis

Despite preceding related publications, works dealing with the resolution of software engineering problems by search techniques has especially risen since 2001. By its first decade, the Search Based Software Engineering (SBSE) approach has been successfully employed in several software engineering contexts, using various optimization techniques. Aside the relevance of such applications, knowledge regarding the publication patterns on the field plays an important role to its understanding and identity. Such information may also shed light into SBSE trends and future. This paper presents the first bibliometric analysis to SBSE publications. The study covered 740 publications of the SBSE community from 2001 through 2010. The performed bibliometric analysis concerned mainly in four categories: Publication, Sources, Authorship, and Collaboration. Additionally, estimates for the next years of several publication metrics are given. The study also analyzed the applicability of bibliometric laws in SBSE, such as Bradfords and Lotka.

The Human Competitiveness of Search Based Software Engineering

This paper reports a comprehensive experimental study regarding the human competitiveness of search based software engineering (SBSE). The experiments were performed over four well-known SBSE problem formulations: next release problem, multi-objective next release problem, workgroup formation problem and the multi-objective test case selection problem. For each of these problems, two instances, with increasing sizes, were synthetically generated and solved by both metaheuristics and human subjects. A total of 63 professional software engineers participated in the experiment by solving some or all problem instances, producing together 128 responses. The comparison analysis strongly suggests that the results generated by search based software engineering can be said to be human competitive.

Automated test case prioritization with reactive GRASP

Modifications in software can affect some functionality that had been working until that point. In order to detect such a problem, the ideal solution would be testing the whole system once again, but there may be insufficient time or resources for this approach. An alternative solution is to order the test cases so that the most beneficial tests are executed first, in such a way only a subset of the test cases can be executed with little lost of effectiveness. Such a technique is known as regression test case prioritization. In this paper, we propose the use of the Reactive GRASP metaheuristic to prioritize test cases. We also compare this metaheuristic with other search-based algorithms previously described in literature. Five programs were used in the experiments. The experimental results demonstrated good coverage performance with some time overhead for the proposed technique. It also demonstrated a high stability of the results generated by the proposed approach.

Parallelizing Feature Selection

AbstractClassification is a key problem in machine learning/data mining. Algorithms for classification have the ability to predict the class of a new instance after having been trained on data representing past experience in classifying instances. However, the presence of a large number of features in training data can hurt the classification capacity of a machine learning algorithm. The Feature Selection problem involves discovering a subset of features such that a classifier built only with this subset would attain predictive accuracy no worse than a classifier built from the entire set of features. Several algorithms have been proposed to solve this problem. In this paper we discuss how parallelism can be used to improve the performance of feature selection algorithms. In particular, we present, discuss and evaluate a coarse-grained parallel version of the feature selection algorithm FortalFS. This algorithm performs well compared with other solutions and it has certain characteristics that makes it a good candidate for parallelization. Our parallel design is based on the master--slave design pattern. Promising results show that this approach is able to achieve near optimum speedups in the context of Amdahls Law.

A pattern language for metadata-based frameworks

Metadata-based frameworks are those that process their logic based on the metadata of the classes whose instances they are working with. Many recent frameworks use this to get a higher reuse level and to be more suitably adapted to the application needs. However, there is not yet a complete best practices documentation or reference architecture for the development of frameworks by using the metadata approach. As a result, this paper presents a pattern language that addresses preliminarily the internal structure of metadata-based frameworks, helping in the understanding and development of such kind of framework.

An ant colony optimization approach to the software release planning with dependent requirements

Ant Colony Optimization (ACO) has been successfully employed to tackle a variety of hard combinatorial optimization problems, including the traveling salesman problem, vehicle routing, sequential ordering and timetabling. ACO, as a swarm intelligence framework, mimics the indirect communication strategy employed by real ants mediated by pheromone trails. Among the several algorithms following the ACO general framework, the Ant Colony System (ACS) has obtained convincing results in a range of problems. In Software Engineering, the effective application of ACO has been very narrow, being restricted to a few sparse problems. This paper expands this applicability, by adapting the ACS algorithm to solve the well-known Software Release Planning problem in the presence of dependent requirements. The evaluation of the proposed approach is performed over 72 synthetic datasets and considered, besides ACO, the Genetic Algorithm and Simulated Annealing. Results are consistent to show the ability of the proposed ACO algorithm to generate more accurate solutions to the Software Release Planning problem when compared to Genetic Algorithm and Simulated Annealing.

STochFS: a framework for combining feature selection outcomes through a stochastic process

The Feature Selection problem involves discovering a subset of features such that a classifier built only with this subset would have better predictive accuracy than a classifier built from the entire set of features. Ensemble methods, such as Bagging and Boosting, have been shown to increase the performance of classifiers to remarkable levels but surprisingly have not been tried in other parts of the classification process. In this paper, we apply the ensemble approach to feature selection by proposing a systematic way of combining various outcomes of a feature selection algorithm. The proposed framework, named STochFS, have been shown empirically to improve the performance of well-known feature selection algorithms.

A Multiobjective Optimization Approach to the Software Release Planning with Undefined Number of Releases and Interdependent Requirements

In software development, release planning is a complex activity which involves several aspects related to which requirements are going to be developed in each release of the system. The planning must meet the customers’ needs and comply with existing constraints. This paper presents an approach based on multiobjective optimization for release planning. The approach tackles formulations when the number of releases is not known a priori and also when the stakeholders have a desired number of releases (target). The optimization model is based on stakeholders’ satisfaction, business value and risk management. Requirements interdependencies are also considered. In order to validate the approach, experiments are carried out and the results indicates the validity of the proposed approach.

A fuzzy approach to requirements prioritization

One of the most important issues in a software development project is the requirements prioritization. This task is used to indicate an order for the implementation of the requirements. This problem has uncertain aspects, therefore Fuzzy Logic concepts can be used to properly represent and tackle the task. The objective of this work is to present a formal framework to aid the decision making in prioritizing requirements in a software development process, including ambiguous and vague data.

Software Next Release Planning Approach through Exact Optimization

Software Requirements phase has notable importance, since it is responsible for the definition of the system itself. Several customers indicate which functionalities they want to be present in the software. However, constraints, such as budget, make it impossible to implement all desired requirements at once. One activity in this context is the release planning. The selection of which requirements should be implemented to the next release is necessary. In literature, metaheuristics have been employed to solve this problem. The objective of this work is to propose the use of exact optimization techniques in the problem, with the advantage that the resolution through these techniques ensures the best solutions. The results in several experiments show the validity of such application, in comparison with the metaheuristics approach.