Topological Deep Learning Model for Thyroid Multi-Class Categorization

Abstract

This study presents an innovative method for categorizing thyroid disorders using a topological deep learning (TDL) model. The strategy utilizes supervised learning and places particular emphasis on topological data analysis (TDA). The work employs a dataset obtained from the WEKA simulator, consisting of 32 characteristics and 3897 instances. It creates a predictive model using the WEKA 3.8.4 simulator, deviating from conventional machine learning methods by giving priority to topological features to improve prediction and classification. The TDL model includes the steps of data preprocessing, selecting unique features, and using TDL algorithms such as Naive Bayes, Decision Trees, and Artificial Neural Networks. The model utilizes embedding to extract information, investigates the importance of specific dimensions, and applies an algorithm to make accurate predictions. The results demonstrate excellent performance, with accuracy, recall, F1-score, and ROC values of 0.96, 0.97, 0.95, and 0.96, respectively. The multi-class classification study provides additional validation of the model's effectiveness, demonstrating an exceptional accuracy of 97.83% and a minimal error rate of 0.21. This highlights the model's precision and dependability, making it suitable for possible clinical applications. This study highlights the efficacy of using topological characteristics and topological data analysis (TDA) methods in the classification of thyroid conditions into many classes. This approach shows potential for enhancing accuracy and dependability in intricate classification tasks.