Benchmarking Generative Adversarial Networks Against Traditional Classifiers for Parkinson’s Disease Detection
Ochiq kirish: yoʻq
Annotatsiya
Parkinson’s disease (PD) is a progressive brain condition that slowly robs people of smooth speech and movement as its dopamine-making nerve cells break down. Catching it early matters, yet doing so is tough because early signs blend with normal aging, and standard tests are far from foolproof. In this work, we put forward a voice-analysis system powered by a Generative Adversarial Network (GAN) and benchmarked it against several classic machine-learning (ML) tools for spotting PD. A publicly available voice database was used, comprising a total of 756 audio files. Noise reduction, feature extraction, and normalization techniques have been applied. Results have been measured based on accuracy, precision, recall, and F1-score for existing algorithms such as Support Vector Machines, Random Forest Classifiers, Logistic Regression, and K-Nearest Neighbors. KNN came out on top, showing a test accuracy of 91.45%, while the GAN-enhanced pipeline still reached a strong 90.13%, and flagged every PD case with zero false negatives, which hints that adding synthetic samples can ease real-world data shortages. The study shows that while machine-learning routines give solid benchmarks, their generative-adversarial-network approach proves far more useful in real hospitals, where catching rare diseases early and gently matters most.
Hali tarjima qilinmagan