Arwan, Achmad (2024) TESLA: Pengembangan Model Tekstual dan Struktural untuk Menemukan Lokasi Fitur. Doctoral thesis, Institut Teknologi Sepuluh Nopember.
Text
7025211031-Dissertation.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (5MB) |
Abstract
Lokasi fitur adalah teknik untuk menemukan area dalam kode yang mengimplementasikan suatu fitur perangkat lunak. Mencari lokasi fitur merupakan pekerjaan yang sulit terutama jika dihadapkan pada perangkat lunak yang sudah berjalan, minim dokumentasi dan memiliki kode yang banyak. Beberapa pendekatan telah dilakukan dalam mencari lokasi fitur yaitu menggunakan data tekstual dengan information retrieval, model. Pendekatan tersebut memiliki kelemahan karena adanya perbedaan istilah antara level tinggi (fitur) dan level rendah (kode sumber). Perangkat lunak dirancang berdasarkan use case scenario sehingga Kueri dengan use case scenario untuk mencari lokasi fitur dapat bermanfaat. Namun, tidak setiap kata dari use case scenario dapat menunjukkan fitur lokasi. Penelitian ini memperkenalkan metode transformasi kata Kueri ke kata yang bersifat teknis yaitu nama Klas, Method dan Javadoc untuk menemukan lokasi fitur pada kode sumber yang sudah diekstraksi struktur hubungannya dan diindeksasi dengan VSM-Lucene. Hasil pencarian kemudian diukur dengan pengukuran precision dan recall. Hasil presisi terbaik adalah 91% dan recall terbaik adalah 97%.
Hasil penelitian ini yaitu berupa: daftar dan karakteristik penelitian lokasi fitur, hasil formulasi metode hibrida (teknik tekstual dan struktural), serta tingkat keberhasilan pengujian dari metode hibrida (teknik tekstual dan struktural).
============================================================
Feature location is a technique for finding areas in code that implement a software feature. Finding feature locations is a difficult task, especially when faced with software that is already running, has minimal documentation and has a lot of code. Several approaches have been taken to find feature locations, namely using textual data with information retrieval models. This approach has weaknesses due to the difference in terms between high level (features) and low level (source code). The software is designed based on use case scenarios so that Kueries with use case scenarios to find feature locations can be useful. However, not every word of the use case scenario can indicate location features. This research introduces a method of transforming query words into technical words, namely Class Name, Method and Javadoc to find the location of features in source code that have had their relationship structure extracted and indexed with VSM-Lucene. The search results are then measured using precision and recall measurements. The best precision result was 91% and the best recall was 97%.
The results of this study are in the form of: a list and characteristics of the feature location research, the results of the hybrid method formulation (textual and structural techniques), as well as the success rate of testing of the hybrid method (textual and structural techniques).
Item Type: | Thesis (Doctoral) |
---|---|
Subjects: | T Technology > T Technology (General) > T58.5 Information technology. IT--Auditing T Technology > T Technology (General) > T59.7 Human-machine systems. |
Divisions: | Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55001-(S3) PhD Thesis (Comp Science) |
Depositing User: | Achmad Arwan |
Date Deposited: | 07 Aug 2024 02:12 |
Last Modified: | 07 Aug 2024 02:12 |
URI: | http://repository.its.ac.id/id/eprint/113842 |
Actions (login required)
View Item |