Detecting Parkinson disease using signals of speech data with ensemble learning

Fig2 from paper “Detecting Parkinson’s disease with sustained phonation and speech signals using machine learning techniques” by Jefferson S.Almeida et al

Decription:

This Project is a group work with Zahra Habibzadeh. in this project we used signals of speech data for a binary classification task to check if the person to whom the sound belongs has Parkinson’s disease or not.

The data used in this project is related to an article with the following title from Jefferson S.Almeida et al:

“Detecting Parkinson’s disease with sustained phonation and speech signals using machine learning techniques”

We implemented the methods used in this paper for data preprocessing, dimensionality reduction and model training. Also, in addition to these cases, we also used ensemble models for training and better performance.

Result:

The best performance was related to the ensemble model using KNN algorithm, whose accuracy was equal to 96,10%. The results of the rest of the models are available in the table below.

Algorithms Accuracy%
Logistic regression 87,23
SVM 93,97
Decision Tree 84,04
KNN(K=1) 94,33
MLP 91,84
RBF 93,26
Ensemble 96,10
Amir Mesbah
Amir Mesbah
Master student in Computer Engineering (with a major in AI)

I am a graduated Master’s student from University of Tehran.