Adopsi Metode Penelusuruan Kebutuhan Pada Lingkungan Agile (Extreme Programming) Untuk Klasifikasi Defect

Hidayati, Nuraisa Novia (2021) Adopsi Metode Penelusuruan Kebutuhan Pada Lingkungan Agile (Extreme Programming) Untuk Klasifikasi Defect. Masters thesis, Institut Teknologi Sepuluh Nopember.

[img] Text
05111950010019-Master_Thesis.pdf - Accepted Version
Restricted to Repository staff only until 1 October 2023.

Download (7MB) | Request a copy

Abstract

Proses pengembangan aplikasi dengan metode agile terutama Extreme Programming (XP) sangat popular saat ini dikarenakan proses iterasi yang singkat. Namun, metode ini memiliki kelemahan pada sisi dokumentasi sehingga penelusuran penerapan kebutuhan menjadi sulit dilakukan. Kontribusi dari penelitian ini adalah memperbaiki penjabaran kebutuhan dan proses penelusuran penerapan kebutuhan sepanjang pengembangan menggunakan metode XP untuk mengenali bagaimana jenis defect major muncul. Pada penelitian ini setiap tahapan penting XP di dokumentasikan serta dilihat keterkaitannya menggunakan Requirement Traceability Matrix (RTM). Setelah RTM terbentuk, kemunculan defect ditelusuri melalui dokumen kebutuhan, pengujian, dan sumber kode pada setiap tahapan sebagai ground truth. Variabel pada ketiga dokumen berkorelasi dengan kemunculan defect. Berdasarkan hasil analisis korelasi Pearson, defect dan bug berkorelasi erat dengan perubahan maupun penambahan kebutuhan, dan tahapan pengujian terutama integration testing. Transformasi vektor diterapkan pada kata di setiap dokumen menggunakan TF-IDF. Hasil vektorisasi menjadi masukan pada tiga metode pemodelan topik, yaitu Latent Semantic Analysis (LSA), Latent Dirichlet Allocation (LDA), dan Non-Negative Matrix Factorization (NMF). Pada dua data set aplikasi metode NMF bekerja dengan baik, meskipun pada aplikasi kedua LSA lebih unggul sedikit dalam nilai akurasi dan recall. Kemunculan term pada setiap topik menjadi atribut klasifikasi defect yang dilakukan paling terakhir dengan menggunakan decision tree dengan dua kelas yaitu defect major dan minor. Hasil akurasi menggunakan decision tree Cart mencapai 94% dengan precision 91% , dan recall 88% dengan empat cabang penting yaitu jumlah pengujian diulang hingga defect teratasi, berapa banyak perubahan kebutuhan yang terjadi pada proses tersebut, konsistensi istilah dari kebutuhan ke pengujian, dan juga dari pengujian ke sumber kode. ====================================================================================================== The application development process using agile methods, especially extreme programming (XP), is prevalent today due to the short iteration process. However, this method has weaknesses on the documentation side, so tracing the application of requirements becomes difficult. The contribution of this research is to improve the description of requirements and the process of tracing the application of requirements throughout development using the XP method to identify how major defects arise. In this study, each important stage of XP is documented, and its relationship is seen using the Requirement Traceability Matrix (RTM). After the RTM is formed, defects are traced through requirements documents, testing, and source code at each stage as ground truth. The variables in the three documents are correlated with the occurrence of defects. Based on the Pearson Correlation Analysis results, defects and bugs are closely correlated with changes or additions to requirements and the testing stages, especially integration testing. Vector transformations are applied to words in each document using TF-IDF. The vectorization results are input for three topic modelling methods, namely Latent Semantic Analysis (LSA), Latent Dirichlet Allocation (LDA), and Non-Negative Matrix Factorization (NMF). In the two application data sets, the NMF method worked well, although, in the second application, LSA was slightly superior in accuracy and recall values. The term's appearance on each topic becomes an attribute of the defect classification that is carried out most recently by using a Decision tree with two classes, namely major and minor defects. The accuracy results using the Decision tree Cart reached 94% with 91% precision, and 88% recall with four important branches, namely the number of tests repeated until the defect was resolved, how many changes in requirements occurred in the process, consistency of terms from requirements to testing, and also from testing to source code.

Item Type: Thesis (Masters)
Uncontrolled Keywords: requirement traceability, RTM, XP, Agile, Information Retrieval, Topic Modelling, TF-Idf, LSA, LDA, NMF, Decision tree, Penelusuran Kebutuhan
Subjects: Q Science > QA Mathematics > QA76.758 Software engineering
Divisions: Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55101-(S2) Master Thesis
Depositing User: Nuraisa Novia Hidayati
Date Deposited: 12 Aug 2021 14:21
Last Modified: 12 Aug 2021 14:21
URI: https://repository.its.ac.id/id/eprint/86019

Actions (login required)

View Item View Item