Pengembangan Prediksi Kerusakan Perangkat Lunak Menggunakan Kombinasi Oversampling dan Undersampling

Iswafaza, Aizul Faiz (2022) Pengembangan Prediksi Kerusakan Perangkat Lunak Menggunakan Kombinasi Oversampling dan Undersampling. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 6025201007-Master_Thesis.pdf Download (2MB)

Abstract

Kualitas perangkat lunak dapat ditingkatkan dengan melakukan pengujian perangkat lunak, akan tetapi semakin banyak fitur yang dikembangkan semakin banyak juga sumber daya yang dibutuhkan. Karena itu prediksi kecacatan perangkat lunak (SDP) diperkenalkan. Akan tetapi dataset yang memunculkan berbagai macam permasalahan seperti redudansi data dan kelas tidak seimbang. Penelitian ini mengusulkan tiga model kombinasi oversampling dan undersampling dalam pengembangan SDP. Model pertama (FCM) merupakan kombinasi dengan alur melakukan undersampling terlelbih dahulu kemudian baru oversampling, model kedua (SCM) merupakan kombinasi dengan alur melakukan oversampling kemudian undersampling dan model (TCM) ketiga penggabungan dari hasil metode oversampling dan undersampling. RSMOTE dan ENN digunakan sebagai metode oversampling dan undersampling pada penelitian ini. Hasil dari FCM, SCM dan TCM memberikan serangkaian dataset baru yang lebih seimbang dan juga bersih dari data ambigu, bising dan duplikasi. Data baru yang dihasilkan model FCM, SCM dan TCM diterapkan pada model prediksinya seperti Artificial Neural Network, Recurrent Neural Network dan Convolutional Neural Network. Kemudian dievaluasi menggunakan pengukuran F-Measure, ROC/AUC dan Akurasi. Dari hasil evaluasi memberikan hasil yang lebih baik dibandingkan dengan penggunaan data original, RSMOTE dan ENN ditunjukkan dengan rata-rata nilai ketika menggunakan model FCM dengan ketiga model prediksi sebesar 0.866, 0.868 dan 0.869. model SCM sebesar 0.855, 0.879 dan 0.882. Dan model TCM sebesar 0.879, 0.864 dan 0.860.
=================================================================================================================================
Software quality can be improved by testing the software, but the more features that are developed, the more resources are needed. Therefore software defect prediction (SDP) was introduced. However, datasets that raise various kinds of problems such as data redundancy and unbalanced classes. This study proposes three combination models of oversampling and undersampling in the development of SDP. The first model (FCM) is a combination with the flow of undersampling first and then oversampling, the second model (SCM) is a combination with the flow of oversampling then undersampling and the third model (TCM) combining the results of the oversampling and undersampling methods. RSMOTE and ENN were used as oversampling and undersampling methods in this study. The results from FCM, SCM and TCM provide a new set of datasets that are more balanced and also clear of ambiguous, noisy and duplication of data. The new data generated by the FCM, SCM and TCM models are applied to their prediction models such as Artificial Neural Network, Recurrent Neural Network and Convolutional Neural Network. Then evaluated using F-Measure, ROC/AUC and Accuracy measurements. From the evaluation results, it gives better results compared to the use of the original data, RSMOTE and ENN indicated by the average value when using the FCM model with the three prediction models of 0.866, 0.868 and 0.869. SCM models are 0.855, 0.879 and 0.882. And the TCM models are 0.879, 0.864 and 0.860.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Kerusakan Perangkat Lunak, Prediksi, Kombinasi Oversampling dan Undersampling, Deep Learning, RSMOTE.ENN, Software Defect, Prediction, Combine of Oversampling and Undersampling
Subjects:	Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) R Medicine > R Medicine (General) > R858 Deep Learning
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55101-(S2) Master Thesis
Depositing User:	Anis Wulandari
Date Deposited:	30 May 2024 07:37
Last Modified:	30 May 2024 07:37
URI:	http://repository.its.ac.id/id/eprint/107999

Actions (login required)

View Item