Earticle

Unsupervised learning techniques such as clustering can be used in software fault prediction, where fault labels are not available. The accurate prediction of faults is likely to occur in coding and that can be rectified early testing phase, which reduces the testing cost as well as maintenance cost and enhance the quality of software. In respect of data mining approach, if training data are not present, then I can not use the supervised learning, this is the biggest problem. To solve this problem, new models using unsupervised learning such as clustering algorithms are quite necessary. This work is the extended version of a Various Strategies and Technical Aspects of Data Mining: A Theoretical Approach, which moves towards practical implementation from theoretical foundation [1]. The main objective of this work to find whether software is faulty or non faulty by using the confusion matrix and also calculating the False Positive Rate (FPR), False Negative Rate (FNR), and Error or fault in a software module. In order to obtain the results I have used an indigenous tool.

Software fault prediction models using supervised learning cannot be applied when training data are not present. In this case, new models using unsupervised learning such as clustering algorithms are quite necessary. Nevertheless, there exist very few studies about unsupervised models because it is difficult to construct the models. One of the difficulties is to decide the number of clusters. To solve this problem, we build unsupervised models using clustering algorithms, EM and Xmeans, which determine the number of clusters automatically and compare them with results of earlier studies. Experimental results show the Xmeans model outperforms the other models.

This paper discuss different classification methods toward reliability and quality improvement of software systems by predicting fault-prone module before testing. Classification capability of Data mining techniques and Object-oriented property based knowledge stored in Object-Oriented metrics are used to classify the software module as fault-prone in different error categories or not fault-prone. Three versions of Eclipse, the java-based Open source Integrated Development environment as dataset for training and testing all the classification based data mining techniques are used. First of all, Threshold base feature ranking (i.e. Area under the ROC curve) is used for selecting effective OO-metrics in building prediction model. After that using those subsets of selected attributes, classification models are built with 41 different classifiers for multinomial classification in fault detection systems. Finally, the performance of a classifier is evaluated with respect to the PRC performance metric. Based on the performance results appropriate classifiers (Random Committee, Random Tree, Randomizable Filtered classifier and IBK) which depict a higher performance and accuracy compared to the others are selected. Our results indicate that Random Tree, Random Committee and Randomizable Filtered Classifier have same performance. IBK classifier also has same performance but little bit less and Kstar has less performance compared to previous four selected classifiers.

Many metric-based classification models have been developed and applied to software fault- proneness prediction. This paper presents a novel prediction model using Random Forest classifier. Random Forest (RF) can be a promising candidate for software quality prediction because it is one of the most accurate classification algorithms available and has strengths in noise handling and efficient running on large data sets. The RF model is constructed and the attribute selection process of the input data is performed before the model evaluation. Prediction accuracy of the model is evaluated using two prediction error measures, Type I and Type II error rates, and compared with well-known prediction models, MultiLayer Perceptron (MLP) neural network model and Support Vector Machine (SVM) model. The results show that the RF model significantly outperforms the SVM model and slightly outperforms the MLP model.

Software plays a significant role in technological and economic development due to its utmost importance in day to day activities. A sequence of rigorous activities under certain constraints is followed to come up with reliable software. Various measures are taken during the process of software development to ensure high quality software. One such method is software module fault prediction for quality assurance to discover defects in the software prior to testing. It aids in predicting the software module faults earlier in the development of the software which predicts fault prone modules so that these can be given special attention to avoid any future risk which eventually curbs the testing along with maintenance cost and effort. The literature survey uncovers many findings that had never been focused like dimensionality reduction and feature selection based on individual feature importance which leads to increase in time complexity and chances of false information. This paper addresses these issues and proposes a supervised machine learning based software module fault prediction technique by implementing Convolutional Neural Network (CNN) as classifier model. Feature selection methods used are InfoGain and Correlation. The results obtained are compared with the existing method HySOM (SOM Clustering with Artificial Neural Network Classification) by considering three different feature sets (Fifteen features, Eighteen features and Twenty one features) of PC1 dataset from NASA. The comparative analysis is performed on the basis of accuracy, precision, recall and F1-measure. The results clearly show better performance of the proposed CNN based technique than HySOM. This paper will contribute towards improvement of quality assurance models utilized for software fault prediction by automating this process using machine learning which enhances True Positive Rate and reduces the detection error. This in turn will help project managers, testers and developers to locate and keep track of fault prone modules so that final software is more accurate, consistent and reliable without consuming much of the testing and maintenance resources.

소프트웨어 결함 예측 연구들의 대부분은 라벨 데이터를 훈련 데이터로 사용하는 감독형 모델에 관한 연구들이 다. 감독형 모델은 높은 예측 성능을 지니지만 대부분 개발 집단들은 충분한 라벨 데이터를 보유하고 있지 않다. 언라벨 데이터만 훈련에 사용하는 비감독형 모델은 모델 구축이 어렵고 성능이 떨어진다. 훈련 데이터로 라벨 데이터와 언라벨 데이터를 모두 사용하는 세미 감독형 모델은 이들의 문제점을 해결한다. Self-training은 세미 감독형 기법들 중 여러 가정과 제약조건들이 가장 적은 기법이다. 본 논문은 Self-training 알고리즘들을 이용해 여러 모델들을 구현하였으며, Accuracy와 AUC를 이용하여 그들을 평가한 결과 YATSI 모델이 가장 좋은 성능을 보였다.

Most studies of software fault prediction have been about supervised learning models that use only labeled training data. Although supervised learning usually shows high prediction performance, most development groups do not have sufficient labeled data. Unsupervised learning models that use only unlabeled data for training are difficult to build and show poor performance. Semi-supervised learning models that use both labeled data and unlabeled data can solve these problems. Self-training technique requires the fewest assumptions and constraints among semi-supervised techniques. In this paper, we implemented several models using self-training algorithms and evaluated them using Accuracy and AUC. As a result, YATSI showed the best performance.

소프트웨어 결함 예측 작업 시 단순히 결함 유무만을 예측하는 이진 분류 모델에 비해 결함의 심각도 범주를 예측하는 다중 분류 모델은 훨씬 유용하게 사용될 수 있다. 소수의 심각도 기반 결함 예측 모델들이 제안되었지만 딥러 닝 기법을 사용한 분류기는 없었다. 본 논문은 3개, 5개의 은닉층을 갖고 은닉층 노드수가 고정된 구조와 변화하는 구조 의 MLP 모델들을 제작하였다. 모델 평가 실험 결과 기존 기계학습 모델들 중 가장 좋은 성능을 보인 MLPs보다 MLP 기반 딥러닝 모델들은 Accuracy와 AUC 모두 유의미하게 더 우수한 성능을 보였다. 특히 노드수 고정 구조에서는 은닉 층수 3, 배치사이즈 32, 노드수 64인 모델 구조가 가장 좋은 성능을 보였다.

In software fault prediction, a multi classification model that predicts the fault severity category of a module can be much more useful than a binary classification model that simply predicts the presence or absence of faults. A small number of severity-based fault prediction models have been proposed, but no classifier using deep learning techniques has been proposed. In this paper, we construct MLP models with 3 or 5 hidden layers, and they have a structure with a fixed or variable number of hidden layer nodes. As a result of the model evaluation experiment, MLP-based deep learning models shows significantly better performance in both Accuracy and AUC than MLPs, which showed the best performance among models that did not use deep learning. In particular, the model structure with 3 hidden layers, 32 batch size, and 64 nodes shows the best performance.

수십년간 매우 많은 소프트웨어 결함 예측 모델에 관한 연구들이 수행되었으며, 그들 중 기계학습 기법을 사용한 모델들이 가장 좋은 성능을 보였다. 딥러닝 기법은 기계학습 분야에서 가장 각광받는 기술이 되었지만 결함 예측 모델의 분류기로 사용된 연구는 거의 없었다. 몇몇 연구들은 모델의 입력 소스나 구문 데이터로부터 시맨틱 정보를 얻어내는데 딥러닝을 사용하였다. 본 논문은 3개 이상의 은닉층을 갖는 MLP를 이용하여 모델 구조와 하이퍼 파라미터를 변경하여 여러 모델들을 제작하였다. 모델 평가 실험 결과 MLP 기반 딥러닝 모델들은 기존 결함 예측 모델들과 Accuracy는 비슷 한 성능을 보였으나 AUC는 유의미하게 더 우수한 성능을 보였다. 또한 또다른 딥러닝 모델인 CNN 모델보다도 더 나은 성능을 보였다.

Many studies have been conducted on software fault prediction models for decades, and the models using machine learning techniques showed the best performance. Deep learning techniques have become the most popular in the field of machine learning, but few studies have used them as classifiers for fault prediction models. Some studies have used deep learning to obtain semantic information from the model input source code or syntactic data. In this paper, we produced several models by changing the model structure and hyperparameters using MLP with three or more hidden layers. As a result of the model evaluation experiment, the MLP-based deep learning models showed similar performance to the existing models in terms of Accuracy, but significantly better in AUC. It also outperformed another deep learning model, the CNN model.

Prediction of fault-prone modules continues to attract researcher's interest due to its significant impact on software development cost. The most important goal of such techniques is to correctly identify the modules where faults are most likely to present in early phases of software development lifecycle. Various software metrics related to modules level fault data have been successfully used for prediction of fault-prone modules. Goal of this research is to predict the faulty modules at design phase using design metrics of modules and faults related to modules. We have analyzed the effect of pre-processing and different machine learning schemes on eleven projects from NASA Metrics Data Program which offers design metrics and its related faults. Using seven machine learning and four preprocessing techniques we confirmed that models built from design metrics are surprisingly good at fault proneness prediction. The result shows that we should choose Naive Bayes or Voting feature intervals with discretization for different data sets as they outperformed out of 28 schemes. Naive Bayes and Voting feature intervals has performed AUC > 0.7 on average of eleven projects. Our proposed framework is effective and can predict an acceptable level of fault at design phases.

An understanding of quality attributes is relevant for the software organization to deliver high software reliability. An empirical assessment of metrics to predict the quality attributes is essential in order to gain insight about the quality of software in the early phases of software development and to ensure corrective actions. In this paper, we predict a model to estimate fault proneness using Object Oriented CK metrics and QMOOD metrics. We apply one statistical method and six machine learning methods to predict the models. The proposed models are validated using dataset collected from Open Source software. The results are analyzed using Area Under the Curve (AUC) obtained from Receiver Operating Characteristics (ROC) analysis. The results show that the model predicted using the random forest and bagging methods outperformed all the other models. Hence, based on these results it is reasonable to claim that quality models have a significant relevance with Object Oriented metrics and that machine learning methods have a comparable performance with statistical methods.