Advanced hybrid feature selection techniques for analyzing the relationship between 25-OHD and TSH

The process of selecting essential features from the high-dimensional datasets is a crucial task while handling biological data. It is not just about choosing the right features but also ensuring that the selected features consistently perform well across different datasets or under varying conditions. The selection of features is crucial for the development of machine learning algorithm, as it influences the model’s characteristics and its connection to physiological processes, which is essential in the healthcare sector for identifying illness states with minimal data. The present study is primarily concerned with addressing the challenge of finding the features which are highly correlated with the TSH (Thyroid Simulating Hormone) from the thyroid datasets, since TSH has been regarded as a primary cause of many ailments. It also tries to find the impact of Vitamin D (25OHD) on TSH. By reducing data dimensions using feature selection techniques, performance and accuracy were improved [1]. Using feature selection algorithms for healthcare problems can reduce diagnosis costs, thereby enhancing healthcare's ability to accurately and promptly identify diseases [2]. This work developed two hybrid Feature Selection Techniques (FST - CorrRecursive Feature Selection (CRFS) and RanChi Ensemble Selection (RCES)), combining the specialties of filter and wrapper methods for identifying the influence of Vitamin D and other features on thyroid. Other existing feature selection methods have also been attempted. The findings demonstrated that, when compared to other approaches, our proposed CRFS and RCES techniques produced superior outcomes.