SRS for Software with Machine Learning Features
Muhammad Hanif Hibatullah, Yani Widyani · 2024
This research focuses on modifying the software requirements specification (SRS) document for software with machine learning (ML) features. With the increasing integration of ML in software, there are specific requirements that are not covered in the existing SRS document format. The current SRS format used by the Informatics Engineering Study Program at Institut Teknologi Bandung (ITB) also faces this limitation. The modification of the SRS document begins with an analysis of the specific ML requirements, including functional and non-functional requirements such as explainability and fairness. The process of modifying the SRS document includes exploring specific ML requirements from various literature, analyzing information items in the SRS document, and modifying the document according to the identified requirements. Finally, a tool is developed as a guide for writing the modified SRS document. The developed tool includes features to facilitate the writing of the SRS document that has been modified to accommodate the requirements related to the development of ML features. The evaluation and validation results show that the modified SRS document can capture and define software requirements with ML features well. The tool testing results show that the developed tool can be used effectively to assist in creating the SRS document.