Peripheral glandular tissue abnormality is a hallmark of thyroid illness. Hyperthyroidism (overactive thyroid) and hypothyroidism (underactive thyroid) those that affect the thyroid the most frequently result from the gland’s abnormally high or insufficient production of hormones. The penalty area of this study was to improve the identification of thyroid disorders by developing a homogeneous machine learning strategy that makes use of feature selection methodologies. When applied to the healthcare industry, machine learning (ML) can greatly improve the accuracy with which jobs like illness risk prediction are performed, thereby benefiting the people in the communities they are implemented in. According to the research studies, there is still a probability of 12% inaccuracy when it comes to the examination of illnesses by doctors. A unique electric fish optimisation (EFO) feature selection perfect is proposed to extract the most significant characteristics, which are strongly contributing, with the goal of reducing the error rate and further improving the presentation. Instead of relying solely on accuracy, we also take into account area under curve (AUC) and other assessment measures because inaccurate predictions for unbalanced datasets might lead to harmful medical outcomes. In this case, this dataset on thyroid disease is sourced from the UCI ML mine. In addition to the EFO model, this research uses s1elect from model (SFM) and tree-based feature selection to identify the top features that contribute to the model, which are then matched with the ML method XGBoost. The proposed EFO-XGBoost model has an absolute highest level of accuracy of 94.66% and a level of precision of 95%.