An efficient dual classification support using ISPCE and IRR-GCBANN techniques for detection of thyroid disease

Abstract

The thyroid gland generates hormones to influence human body metabolism. A proper analysis of thyroid glands functional data is needed for Thyroid Disease (TD) diagnosis. Detecting the TD early is a crucial issue. Disparate alternatives approach was generated for it over time, but inaccurate detection of TD remains. Thus, aimed at detecting TD, the work renders an effectual dual classification framework. The framework was split into seven phases. Initially, to convert the string data into a numerical value, data numeralization is performed. After that, preprocessing is executed, which processes the missing value by means of taking the average and eliminating the repeated data. Next, so as to scale the entire data into a single unit, normalization is carried out. And the essential features are extracted, and Modified Discrete Salp Swarm Algorithm (MDSSA) Features Selection (FS) technique is developed in order to get rid of relevant features. Currently, an Improved Sign Preserving Cluster Ensemble (ISPCE) is proposed to cluster the chosen features accurately. It clusters the TD utilizing base clusters and lessens the decision graph complications, time intricacy, etc. Finally, the clustered features are inputted to Improved Raven Roosting Optimization Algorithm with Gradient Cats Boost Artificial Neural Network (IRR-GCBANN) for classifying the TD as hypothyroidism, hyperthyroidism, or normal. Experimental outcomes exhibit that the proposed framework attains 96.94% accuracy for detecting the TD when weighted against the existent techniques.