Intelligent Requirements Engineering: Applying Machine Learning for Requirements Classification

The classification of requirements plays a crucial role in requirements engineering, enabling the differentiation between legally relevant requirements and auxiliary content. However, the manual labeling of each content element in a requirements specification as a “functional requirement” or “non-functional requirement” or “information” is a time-consuming and error-prone task. In this paper, we propose an approach that automates the classification of content elements in a natural language requirements specification as either “functional requirement” or “non-functional requirement” or “information” Our approach leverages a combination of Convolutional Neural Networks (CNN) and Support Vector Machines (SVM) for the classification task. The CNN model is responsible for extracting meaningful features from the textual content, while the SVM classifier is employed to make the final classification decision. To train and validate our model, we utilized online datasets specifically designed for requirements classification. Additionally, we augmented these datasets by incorporating data from other projects. The performance of our model was measured using various evaluation metrics, including accuracy, F1 score, precision, recall, and confusion matrix analysis. The experimental results demonstrate promising performance with a precision of 80%, accuracy of 85%, F1 score of 88%, and recall of 90%. These results indicate that our approach successfully automates the classification process and significantly reduces the need for manual labeling, thereby saving time and reducing the potential for errors in requirements classification.