INTEGRASI METODE RESAMPLING DAN K-NEAREST NEIGHBOR PADA PREDIKSI CACAT SOFTWAR APLIKASI ANDROID

Authors

  • Rimbun Siringo-ringo

Abstract

Efects are actual errors that can be found in the software. The existence of defects can make the software does not work optimally even crashes. The existence of defect is something that must be eliminated in order to produce high-quality software. Testing Techniques conventional software is cracking failure is flawed searches on a case by case basis. This technique is done by flawed search case by case basis.  To ensure the quality of the software, we need models and effectively defect testing methods.Method of k-Nearest Neighbor (k-NN) is one of  the popular classification  method that is  widely applied to build  predictive models.  K-NN method is very susceptible to errors caused by the imbalance of the class., because it results in a low level of accuracy. class imbalance produces a low level of accuracy. To address the class imbalance, resampling methods applied at the pre-processing stage. Experiments done by comparing the results that obtained with and without resampling method. To validate the superiority of the model, the experimen results compared to other classification methods, namely J48 and decision table. The results showed that the use of resampling methods produce significant performance improvements to the k-NN. The proposed model is better than the method J48 and decision table.

Keywords : k-Nearest Neighbor, resampling, software defects, android aplication 

Downloads

Published

2017-02-20