Genetic Programming

Abstract

Geometric Semantic Genetic Programming (GSGP) is a novel approach in artificial intelligence that has shown superior performance over traditional methods across various domains. The study explores the capabilities of GSGP within the demanding field of Parkinson’s disease (PD) diagno- sis. This prevalent neurodegenerative disorder poses substantial diagnostics challenges due to the tedious diagnosis process, and its complex symptomatology. The project contributes to PD diag- nostics by leveraging the Unified Parkinson’s Disease Rating Scale (UPDRS), a recognized standard for assessing PD’s severity. The report summarizes relevant literature and details the project spec- ification, and the experimental design and implementation. It discusses the comparative analysis of GSGP against the standard genetic programming (STGP) and 5 other machine learning models across different data feature sets. The study evaluates GSGP’s prediction accuracy, computational efficiency, and stability. It compares these attributes within used models and past applications. Results suggest that GSGP surpasses STGP and outperforms many of the conventional machine learning methods. Despite its promising performance, the study identifies significant potential for further enhancements. Particularly, in terms of hyperparameter settings and extensions in the GSGP technique. The study demonstrates GSGP’s use in complex medical domains and con- tributes to the broader understanding of its potential within such fields. It advocates for continued exploration to fully harness GSGP’s capabilities.