Open AccessDermatology disease prediction based on firefly optimization of ANFIS classifier
Author(s) -
J. Rajeshwari,
M. Sughasiny
Publication year - 2022
Publication title -
aims electronics and electrical engineering
Language(s) - English
Resource type - Journals
ISSN - 2578-1588
DOI - 10.3934/electreng.2022005
Subject(s) - adaptive neuro fuzzy inference system , artificial intelligence , computer science , fuzzy logic , classifier (uml) , feature selection , python (programming language) , firefly algorithm , pattern recognition (psychology) , artificial neural network , machine learning , fuzzy control system , particle swarm optimization , operating system
The rate of increase in skin cancer incidences has become worrying in recent decades. This is because of constraints like eventual draining of ozone levels, air's defensive channel capacity and progressive arrival of Sun-oriented UV radiation to the Earth's surface. The failure to diagnose skin cancer early is one of the leading causes of death from the disease. Manual detection processes consume more time well as not accurate, so the researchers focus on developing an automated disease classification method. In this paper, an automated skin cancer classification is achieved using an adaptive neuro-fuzzy inference system (ANFIS). A hybrid feature selection technique was developed to choose relevant feature subspace from the dermatology dataset. ANFIS analyses the dataset to give an effective outcome. ANFIS acts as both fuzzy and neural network operations. The input is converted into a fuzzy value using the Gaussian membership function. The optimal set of variables for the Membership Function (MF) is generated with the help of the firefly optimization algorithm (FA). FA is a new and strong meta-heuristic algorithm for solving nonlinear problems. The proposed method is designed and validated in the Python tool. The proposed method gives 99% accuracy and a 0.1% false-positive rate. In addition, the proposed method outcome is compared to other existing methods like improved fuzzy model (IFM), fuzzy model (FM), random forest (RF), and Naive Byes (NB).