INTEGRASI TEKNIK SMOTE BAGGING DENGAN INFORMATION GAIN PADA NAIVE BAYES UNTUK PREDIKSI CACAT SOFTWARE
Abstract: The prediction
accuracy of defects in code, can help direct the test effort, reduce costs and
improve software quality. Until now, many researchers have applied various
types of algorithm based on machine learning and statistical methods to build
predictive performance software defects. One of them uses machine learning
approach to the classification, which is a popular approach to predict software
defects. While Naive Bayes one simple classification to have good performance
that produces an average probability of 71 percent. As well as the time
required in the process of learning faster than on any other machine learning.
Additionally it has a good reputation on the accuracy of the prediction. While
NASA MDP is a very popular data used by previous researchers in the development
of predictive models of software defects. Because it is common and freely used
by researchers. However, these data have deficiencies, including the occurrence
of imbalance class and attribute noise. Therefore by using SMOTE (Synthetic
Minority Over-Sampling Technique) for sampling techniques and Bagging on the
ensemble method, is used to deal with the class imbalance. As for dealing with
noise attribute, in this research using information gain in the process of
selecting the relevant attributes. So after the trial that the application of
the model SMOTE Bagging and Information Gain proven to obtain good results to
handled imbalance class and attribute noise at prediction software defects, and
can increase the accuracy of the prediction results software defects.
Penulis: Sukmawati Anggraeni
Putri
Kode Jurnal: jptkomputerdd170253
