Prediction of Thyroid Using Three-Stage Rule-Generative Associative Learning Model

Predicting thyroid disorders has become a significant task in recent times. Even though there are methods for diagnosing it, the commonly employed technique is binary classification. This technique uses tiny datasets, and the outcomes are not verified. Current methods mainly concentrate on optimizing models, with less attention paid to the feature engineering aspect. This work offers a process that looks into feature engineering for DL and ML techniques to overcome these restrictions. Some adopted methods include machine learning-based selecting features with additional tree classifiers, forward selection of features, backward feature removal, and multimodal feature elimination. The suggested method can identify the following conditions: enhanced binding protein, Hashimoto's thyroiditis (main hypothyroid), autoimmune thyroiditis, and non-thyroidal syndrome. Various evaluations reveal that combined with the Three-Stage Rule-Generative Associative Learning Model (TS-RGALM) based classifier selected feature generates the best outcomes by achieving 99% accuracy. The outcomes indicate that machine-learning techniques are more appropriate for thyroid illness identification. The efficacy of the anticipated method is confirmed and comparison with prior research.