Fig 1.
The image illustrates a workflow for processing comments and identifying classifying Self-admitted technical debt.
Fig 2.
Illustrates Workflow for identifying and classifying bugs based on the basis Self-Admitted technical debt.
Fig 3.
The image illustrates a workflow of identifying and classifying bugs based on self admitted technical debt using transfer learning.
Fig 4.
The image Illustrate Workflow for preparing the dataset for identifying and classifying SATD.
Fig 5.
The image Illustrate Distribution of Self-Admitted Technical Debt’s types.
Fig 6.
The image Illustrate Count of instances for each project category.
Fig 7.
The image Illustrate distribution of Text Length Across Different Types of SATD.
Fig 8.
The image Illustrate distribution of different bug components.
Fig 9.
The image Illustrate Visualization of most common terms to understand stop words in short desc.
Table 1.
Class distribution of different types of technical debt.
Fig 10.
The image Illustrate missing values in SATD dataset.
Fig 11.
The image Illustrate class distribution of different types of SATD after balancing the dataset using SMOTE-TEXT technique.
Table 2.
Class distribution of different bug components in the dataset.
Fig 12.
The image Illustrate class distribution of different bug components after balancing the dataset using SMOTE-TEXT technique.
Table 3.
Training Parameters Setting for Baseline, Deep Learning, and Transformer Models on SATD Dataset.
Table 4.
Training parameters setting for pre-trained Baseline, Deep Learning, and transformer Models on Bug Dataset.
Fig 13.
The image Illustrate Transfer learning process for deep learning models.
Fig 14.
The image Illustrated Transfer learning process for Transformer Model.
Table 5.
Detail of precision, recall, accuracy, and F1 Score for Naïve Bayes on SATD.
Table 6.
Detail of precision, recall, accuracy, and F1 Score for AdaBoost Classifier on SATD.
Fig 15.
The image Illustrated the Confusion Matrix provides a detailed summary of the performance of a naïve Bayes model by presenting the counts of true positives (TP), true negatives (TN), false positives (FP), and false negatives (FN).
The rows correspond to the actual classes, while the columns represent the predicted classes. Correct predictions are reflected along the diagonal elements, whereas the off-diagonal elements indicate the instances of misclassification.
Fig 16.
The image Illustrated the Confusion Matrix displays the quantities of true positives (TP), true negatives (TN), false positives (FP), and false negatives (FN) generated by the AdaBoost Classifier model.
The rows denote the actual classes, whereas the columns indicate the predicted classes. The diagonal elements represent the count of accurate predictions, whereas the off-diagonal elements reflect instances of misclassification.
Fig 17.
The image Illustrated the Confusion Matrix displays the quantities of true positives (TP), true negatives (TN), false positives (FP), and false negatives (FN) predicted by the LSTM.
The rows denote the actual classes, whilst the columns signify the predicted classes. The diagonal elements represent the quantity of accurate predictions, and the off-diagonal elements denote misclassifications.
Fig 18.
The image Illustrated the Confusion Matrix displays the quantities of true positives (TP), true negatives (TN), false positives (FP), and false negatives (FN) predicted by the LSTM.
The rows denote the actual classes, whilst the columns signify the predicted classes. The diagonal elements represent the quantity of accurate predictions, and the off-diagonal elements denote misclassifications.
Fig 19.
The image Illustrated the Confusion Matrix displays the quantities of true positive (TP), true negative (TN), false positive (FP), and false negative (FN) predictions generated by the GRU.
The rows denote the actual classes, whereas the columns indicate the predicted classes. The diagonal elements represent the count of accurate predictions, whereas the off-diagonal elements illustrate instances of misclassification.
Fig 20.
The image Illustrated the Confusion Matrix displays the quantities of true positives (TP), true negatives (TN), false positives (FP), and false negatives (FN) predicted by the BI-GRU.
The rows denote the actual classes, whilst the columns signify the predicted classes. The diagonal elements represent the count of accurate predictions, and the off-diagonal elements denote misclassifications.
Table 7.
Detail of precision, recall, accuracy, and F1 Score for LSTM on SATD.
Table 8.
Detail of precision, recall, accuracy, and F1 Score for Bi-LSTM on SATD.
Table 9.
Detail of precision, recall, accuracy, and F1 Score for GRU on SATD.
Table 10.
Detail of precision, recall, accuracy, and F1 Score for Bi-GRU on SATD.
Table 11.
Detail of precision, recall, accuracy, and F1 Score for BERT on SATD.
Table 12.
Detail of precision, recall, accuracy and F1 Score for GPT-3 on SATD.
Fig 21.
The image Illustrated the Confusion Matrix displays the counts of true positives (TP), true negatives (TN), false positives (FP), and false negatives (FN) predicted by BERT.
The rows denote the actual classes, whilst the columns signify the predicted classes. The diagonal elements represent the count of accurate predictions, and the off-diagonal elements denote misclassifications.
Fig 22.
The image Illustrated the Confusion Matrix displays the quantities of true positives (TP), true negatives (TN), false positives (FP), and false negatives (FN) predicted by GPT-3.
The rows denote the actual classes, whilst the columns indicate the anticipated classes. The diagonal elements represent the count of accurate predictions, and the off-diagonal elements denote misclassifications.
Fig 23.
The image Illustrated Comparison of different models trained and evaluated on SATD.
Table 13.
Comparative analysis of different models used for self-admitted technical identification and classification.
Fig 24.
The image Illustrated the Confusion Matrix displays the quantities of true positives (TP), true negatives (TN), false positives (FP), and false negatives (FN) predicted by the pre-trained Naive Bayes model.
The rows denote the actual classes, whilst the columns signify the predicted classes. The diagonal elements represent the count of accurate predictions, and the off-diagonal elements denote misclassifications.
Table 14.
Detail of precision, recall, accuracy and F1 Score for pre-trained Naïve Bayes on Bug Dataset.
Fig 25.
The image Illustrated the Confusion Matrix displays the quantities of true positives (TP), true negatives (TN), false positives (FP), and false negatives (FN) predicted by the Pre-trained LSTM.
The rows denote the actual classes, whilst the columns signify the predicted classes. The diagonal elements represent the count of accurate predictions, and the off-diagonal elements denote misclassifications.
Table 15.
Detail of precision, recall, accuracy and F1 Score for pre-trained LSTM on Bug Dataset.
Fig 26.
The image Illustrated the Confusion Matrix displays the quantities of true positives (TP), true negatives (TN), false positives (FP), and false negatives (FN) predicted by the Pre-trained GRU.
The rows denote the actual classes, whilst the columns signify the predicted classes. The diagonal elements represent the count of accurate predictions, and the off-diagonal elements denote misclassifications.
Table 16.
Detail of precision, recall, accuracy and F1 Score for pre-trained GRU on Bug Dataset.
Fig 27.
The image Illustrated the Confusion Matrix displays the quantities of true positives (TP), true negatives (TN), false positives (FP), and false negatives (FN) predicted by the Pre-trained GPT-3.
The rows denote the actual classes, whilst the columns signify the predicted classes. The diagonal elements represent the count of accurate predictions, and the off-diagonal elements denote misclassifications.
Table 17.
Detail of precision, recall, accuracy and F1 Score for pre-trained GPT-3 on Bug Dataset.
Fig 28.
The image Illustrated Comparison of different models trained and evaluated on Bug Dataset.
Table 18.
Comparative analysis of different models used for bug identification and classification.