Hybrid Genetic Algorithm based Fuzzy Inference System for Data Regression

Abstract

Regression analysis is one of the most popular methods of estimation or forecasting. For someone who is the non-domain expert to understand how the estimation decision is made, clarity and transparency of the regression model is required to reveal knowledge and information that evaluates the functional relationship between two objects, i.e., the independent and dependent objects the system represents. Hence, this paper presents the hybridization of Genetic Algorithm (GA) and Fuzzy Inference System (FIS)-based computational intelligence systems for tackling data regression problem (hereinafter denoted as GA-FIS-RG). With this regard, GA-FIS-RG first defines the membership functions with logical interpretation which is amendable by domain experts to human understanding, and then GA serves as an optimization tool to construct the best combination of rules in fuzzy inference system. For performance evaluations, we demonstrate the interpretability and applicability of GA-FIS-RG to data regression problems, i.e., the Santa-Fe Series-E and Auto MPG.