Didar-Hussain/MSAR-CNN-EARS: MSR-CNN v1.0.0

Didar-Hussain

MSR-CNN: Multi-Scale Residual Convolutional Neural Network for Automated Multiclass Classification of EARS Functional Requirements This repository contains the implementation of the proposed Multi-Scale Residual Convolutional Neural Network (MSR-CNN) developed for the automated multiclass classification of Easy Approach to Requirements Syntax (EARS) functional requirements. The proposed deep learning framework integrates trainable word embeddings, parallel multi-scale convolutional feature extraction, residual learning, batch normalization, and spatial dropout to learn contextual semantic representations directly from textual software requirements. The repository provides the source code, trained model, dataset, and supporting files required to reproduce the experimental results reported in the associated research paper. The proposed MSR-CNN framework was evaluated using seven complementary performance metrics, including Accuracy, Macro Precision, Macro Recall, Macro F1-score, Macro ROC-AUC, Matthews Correlation Coefficient (MCC), and Geometric Mean (G-Mean). Comparative experiments against traditional machine learning classifiers, an ablation study, and a comparison with a recent state-of-the-art method demonstrate the effectiveness and robustness of the proposed architecture for automated EARS functional requirement classification.