Classification of Non-Functional Requirements From IoT Oriented Healthcare Requirement Document.

Area of Science:

• Computer Science
• Artificial Intelligence
• Healthcare Informatics

Background:

• The Internet of Things (IoT) enables smart healthcare systems for patient monitoring and emergency response.
• Machine learning (ML) is used for security in smart healthcare, but classifying non-functional requirements (NFRs) from documents is overlooked.
• Manual NFR classification is error-prone and time-consuming, potentially compromising IoT healthcare system security and performance.

Purpose of the Study:

• To classify non-functional requirements (NFRs) from requirement documents for IoT-oriented healthcare systems.
• To address the gap in automated NFR classification within the Requirement Engineering (RE) phase for healthcare IoT.

Main Methods:

• An experiment was conducted to classify NFRs using various ML algorithms: Logistic Regression (LR), Support Vector Machine (SVM), Multinomial Naive Bayes (MNB), K-Nearest Neighbors (KNN), ensemble, and Random Forest (RF).
• A novel hybrid KNN rule-based ML algorithm was developed and tested for NFR classification.
• A new dataset of 104 requirements specific to IoT-oriented healthcare systems was created.

Main Results:

• The proposed hybrid KNN rule-based ML algorithm achieved the highest average classification accuracy of 75.9%.
• This hybrid approach outperformed standard ML algorithms in classifying NFRs from IoT healthcare requirement documents.

Conclusions:

• The research introduces a novel ML approach and a hybrid KNN rule-based algorithm for classifying NFRs in IoT healthcare systems.
• The developed hybrid algorithm demonstrates superior accuracy in identifying NFRs, enhancing the RE process.
• The creation of a specific dataset aids future research, though its small size may impact generalizability.