Danielle Azar

An ant colony optimization algorithm to improve software quality prediction models

ContextAssessing software quality at the early stages of the design and development process is very difficult since most of the software quality characteristics are not directly measurable. Nonetheless, they can be derived from other measurable attributes. For this purpose, software quality prediction models have been extensively used. However, building accurate prediction models is hard due to the lack of data in the domain of software engineering. As a result, the prediction models built on one data set show a significant deterioration of their accuracy when they are used to classify new, unseen data. ObjectiveThe objective of this paper is to present an approach that optimizes the accuracy of software quality predictive models when used to classify new data. MethodThis paper presents an adaptive approach that takes already built predictive models and adapts them (one at a time) to new data. We use an ant colony optimization algorithm in the adaptation process. The approach is validated on stability of classes in object-oriented software systems and can easily be used for any other software quality characteristic. It can also be easily extended to work with software quality predictive problems involving more than two classification labels. ResultsResults show that our approach out-performs the machine learning algorithm C4.5 as well as random guessing. It also preserves the expressiveness of the models which provide not only the classification label but also guidelines to attain it. ConclusionOur approach is an adaptive one that can be seen as taking predictive models that have already been built from common domain data and adapting them to context-specific data. This is suitable for the domain of software quality since the data is very scarce and hence predictive models built from one data set is hard to generalize and reuse on new data.

A hybrid heuristic approach to optimize rule-based software quality estimation models

Software quality is defined as the degree to which a software component or system meets specified requirements and specifications. Assessing software quality in the early stages of design and development is crucial as it helps reduce effort, time and money. However, the task is difficult since most software quality characteristics (such as maintainability, reliability and reusability) cannot be directly and objectively measured before the software product is deployed and used for a certain period of time. Nonetheless, these software quality characteristics can be predicted from other measurable software quality attributes such as complexity and inheritance. Many metrics have been proposed for this purpose. In this context, we speak of estimating software quality characteristics from measurable attributes. For this purpose, software quality estimation models have been widely used. These take different forms: statistical models, rule-based models and decision trees. However, data used to build such models is scarce in the domain of software quality. As a result, the accuracy of the built estimation models deteriorates when they are used to predict the quality of new software components. In this paper, we propose a search-based software engineering approach to improve the prediction accuracy of software quality estimation models by adapting them to new unseen software products. The method has been implemented and favorable result comparisons are reported in this work.

A GENETIC ALGORITHM FOR IMPROVING ACCURACY OF SOFTWARE QUALITY PREDICTIVE MODELS: A SEARCH-BASED SOFTWARE ENGINEERING APPROACH

In this work, we present a genetic algorithm to optimize predictive models used to estimate software quality characteristics. Software quality assessment is crucial in the software development field since it helps reduce cost, time and effort. However, software quality characteristics cannot be directly measured but they can be estimated based on other measurable software attributes (such as coupling, size and complexity). Software quality estimation models establish a relationship between the unmeasurable characteristics and the measurable attributes. However, these models are hard to generalize and reuse on new, unseen software as their accuracy deteriorates significantly. In this paper, we present a genetic algorithm that adapts such models to new data. We give empirical evidence illustrating that our approach out-beats the machine learning algorithm C4.5 and random guess.

Computer-aided detection of Melanoma using geometric features

Melanoma is one type of skin cancer that usually develops from prolonged exposure to UV light. The latter triggers mutations that lead skin cells to multiply rapidly and form malignant tumors. If not cured, Melanoma can result in ones death. Hence, an early detection of this deadly cancer is important to prevent it. Certain lesion characteristics such as Asymmetry, Border, Color and Diameter segmentation (ABCD rule), can indicate the presence of Melanoma. In this work, we investigate the use of geometric features to differentiate between a benign lesion and a malignant one. The k-Nearest Neighbors (k-NN) machine learning algorithm is used to classify 15 lesions based on their ABD features. An accuracy of 89% was obtained on the testing set. The results indicate that this technique may be used to detect Melanoma skin cancer.

Predicting stability of classes in an object-oriented system

The stability of a class in object-oriented system is one software quality characteristic that is important to assess at the early development stages. However, a direct measure of this software quality characteristic is not possible. Nonetheless, it can be predicted based on other measurable software attributes such as cohesion, coupling, and complexity. Many metrics have been proposed to assess these software attributes and for this purpose, prediction models have been widely used. However, in almost all cases, these models were not efficient when used to predict the quality characteristics (stability or other) of new unseen software as their prediction accuracy decreases significantly. In this paper, we present a heuristic approach that relies on the adaptation and recombination of already built predictive models to new unseen software.The predictive models are all rule-based models and the approach is tested on the stability of classes in an object-oriented software system. We compare our results to the machine learning algorithm C4.5, and we show that our approach out-beats it.

A PSO-GA Approach Targeting Fault-Prone Software Modules

Abstract We present an algorithm to classify software modules as fault-prone or not using object-oriented metrics. Our algorithm is a combination of particle swarm intelligence and genetic algorithms. We empirically validate it on eight different data sets. We also compare it to well known classification techniques. Results show that our algorithm has several advantages over other techniques.

An Image Processing and Genetic Algorithm-based Approach for the Detection of Melanoma in Patients

Melanoma skin cancer is the most aggressive type of skin cancer. It is most commonly caused by excessive exposure to Ultraviolet radiation which triggers uncontrollable proliferation of melanocytes. Early detection makes melanoma relatively easily curable. Diagnosis is usually done using traditional methods such as dermoscopy which consists of a manual examination performed by the physician. However, these methods are not always well founded because they depend heavily on the physicians experience. Hence, there is a great need for a new automated approach in order to make diagnosis more reliable. In this paper, we present a twophase technique to classify images of lesions into benign or malignant. The first phase consists of an image processing-based method that extracts the Asymmetry, Border Irregularity, Color Variation and Diameter of a given mole. The second phase classifies lesions using a Genetic Algorithm. Our technique shows a significant improvement over other well-known algorithms and proves to be more stable on both training and testing data.

Heuristic approaches for optimizing the performance of rule-based classifiers

Rule-based classifiers are supervised learning techniques that are extensively used in various domains. This type of classifiers is popular because of its nature which makes it modular and easy to interpret and also because of its ability to provide the classification label as well as the reason behind it. Rule-based classifiers suffer from a degradation of their accuracy when they are used on new data. In this paper, we present an approach that optimizes the performance of the rule-based classifiers on the testing set. The approach is implemented using five different heuristics. We compare the behavior on different data sets that are extracted from different domains. Favorable results are reported.