A new approach due to gender based on hybrid machine learning for diagnosis of parkinson's disease from sound signals

Parkinson's disease causes disruption in many vital functions such as speech, walking, sleeping, and movement, which are the basic functions of a human being. Early diagnosis is very important for the treatment of this disease. In order to diagnose Parkinson's disease, doctors need brain tomography, and some biochemical and physical tests. In addition, the majority of those suffering from this disease are over 60 years of age, make it difficult to carry out the tests necessary for the diagnosis of the disease. This difficult process of diagnosing Parkinson's disease triggers new researches. In this thesis, it is aimed to diagnose Parkinson's disease with models based on hybrid machine learning with the help of acoustic sounds. For this purpose, 188 (107 Male -81 Female) individuals with Parkinson's disease and 64 healthy (23 Male - 41 Female) individuals were asked to say the letter "a" three times and their measurements were made and recorded. In this study, the data set of recorded 756 measurements was used. By separating this data set based on gender (Mr.-Mrs.), three separate data sets (Mr., Mrs. and Mixed) were created. All the next steps are processed separately for these three data sets. First, the audio recordings were devoted to the Baseline, Time, Vocal, MFCC and Wavelet feature groups. Later, data sets were balanced in terms of "Patient / Healthy" feature. Then, with the help of Eta correlation coefficient based feature selection algorithm (E-Score), the best 20% percent feature was selected for each property group. Later, new approaches based on hybrid machine learning were developed using decision trees and support vector machines. For this, data sets are divided into two groups as 75% education and 25% test groups with the help of systematic sampling method. Performance of these designed models were calculated with Accuracy rate, Specificity, Sensitivity, F-Measurement, AUC and Kapa values. In this thesis study; The best accuracy rate, specificity and sensitivity values obtained with the male data set are 93,81%, 1 and 1 respectively. These values obtained with the female data set are 91,21%, 0,97 and 0,98 respectively. Similarly, these values obtained with the raw (mixed) data set are 93,12%, 0,94 and 0,98 respectively. The high success rates obtained show that the models which designed can be used for the diagnosis of Parkinson's disease. In addition, it was observed that the use of gender-based data sets increased performance in the diagnosis of Parkinson's diseas.