MGGP

MGGP applies multiobjective grammar-based genetic programming to model complex multifactorial relationships in epidemiological data, enabling analysis of factors underlying diseases such as asthma and allergies.

Key Features:

• Advanced Machine Learning Approach: MGGP employs a multiobjective grammar-based genetic programming framework to explore interactions among environmental, psychosocial, socioeconomic, nutritional, and infectious factors and capture non-linear relationships.
• Interpretability: MGGP generates models that balance predictive accuracy and interpretability by producing human-readable model structures constrained by grammar rules.
• Comparative Performance: In a study of 1,047 subjects, MGGP outperformed logistic regression and C4.5 decision trees in modeling asthma prevalence and allergy markers (IgE antibody presence and skin prick test positivity) and achieved accuracy comparable to Random Forests while providing greater interpretability.
• Epidemiological Application: MGGP has been applied to pediatric datasets to elucidate how infections, psychosocial stressors, nutrition, hygiene, and socioeconomic status interact to influence asthma and allergy outcomes.

Scientific Applications:

• Disease Modeling: Modeling multifactorial disease etiology for asthma and allergies by capturing nuanced interactions among heterogeneous factors.
• Public Health Research: Disentangling multifactorial influences on disease prevalence in epidemiological and public health studies.

Methodology:

MGGP applies genetic programming within a multiobjective optimization framework, evolving populations of models over successive generations guided by fitness functions that evaluate accuracy and interpretability and enforcing grammar-based rules to constrain model structure.