Improved Multiobjective Multifactor Dimensionality Reduction using Fuzzy Theory

Epistasis identification is essential for understanding the susceptibility of genetic diseases and remains a challenge. Currently, multifactor dimensionality reduction (MDR) has been enhanced using a multiobjective approach to identify epistasis. To reduce the multifactor dimension. MDR-based approaches use a binary classification method to distinguish high risk (H) and low risk (L) groups. However, the limitation of NIDR-based approaches use a binary-classification binary classification that does not reflect the uncertainty of the H/L classification. In this study, we used the membership degree of the empirical fuzzy method to propose an improved multiobjective multifactor dimensionality reduction (MOMDR) approach to overcome the limitation of binary classification. The empirical fuzzy MOMDR considers two fuzzy-based metrics simultaneously, namely the correct classification and likelihood, and does not require adjustment of the parameter values. In simulation studies, the proposed empirical fuzzy MOMDR has demonstrated a higher detection success rate than MDR and MOMDR.