11 III March 2023

https://doi.org/10.22214/ijraset.2023.49497
 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.538
Volume 11 Issue III Mar 2023- Available at www.ijraset.com

A Review on Detection of Parkinsons Disease Using

ML Algorithms
Sakshi M1, Dr. Sankhya N Nayak2, Skanda G N3, Shreyas R Adiga4, Pavana R5
1, 3, 4, 5
Research Scholar, Dept. of CS&E, JNNCE, Shimoga, India
2
Associate Prof., Dept. of CS&E , JNNCE, Shimoga, India

Abstract: Parkinson disease prediction is an area of active research in healthcare and machine learning. Even though
Parkinson's disease is not well-known worldwide, its negative impacts are detrimental and should be seriously considered.
Furthermore, because individuals are so immersed in their busy lives, they frequently disregard the early signs of this condition,
which could worsen as it progresses. There are many techniques for Parkinson disease prediction. In this paper we are going to
discuss some of the possible technical solutions proposed by researchers.
Keywords: Parkinson disease

I. INTRODUCTION
The brain and nervous system are both affected by Parkinson's disease, which is a neurodegenerative condition. The loss of
dopamine-producing neurons in the basic ganglia is specifically related to it. The illness has negative effects on people, society, and
money on a social, professional, and personal level.
Individual symptoms that develop over time and vary from person to person can be divided into two categories. Motor symptoms
include stiffness, slowness, also known as bradykinesia, facial expression, fewer swings of the arms, and resting tremor, whereas
non-motor symptoms, which affect every system and component of the body, are unseen symptoms. These symptoms of autonomic
dysfunction include perspiration, urination, and mood and thought disturbances.
The primary objective of the study is to evaluate the effectiveness of various Supervised Algorithms for enhancing Parkinson
Disease detection diagnosis. Parkinson Disease was predicted using K-Nearest Neighbor, Logistic Regression, Decision Tree, Naive
Bayes, and XGBoost. The detection of Parkinson's disease is based on the use of different classifiers, such as Accuracy, F1-score,
Recall, Precision, R2-score Total UPDRS Motor UPDRS and Confusion matrix.
Amreen Khanum at el. [1] examined the effects of the various Supervised ML Algorithms for upgrading the diagnosis of Parkinson
Disease. KNN, LR, DT, NB, and XGBoost were five machine learning techniques used to detect Parkinson's disease. The
performance of the classifiers was assessed using precision, accuracy, F1-Score, and recall. Data on Parkinson's disease was
obtained for this study from the UCI Machine Learning Repository. 23 speech feature sets are included in the 195 patient records
that constituted this dataset. The first step was the extraction of characteristics from datasets related to Parkinson's disease. The
study used five supervised learning algorithms to recognise Parkinson's illness. As a result, the performance metrics were evaluated
to find the algorithm that outperformed.
Muhtasim Shafi Kader at el.[2] Mushtasim Shafi Kader at the el. [] chose 195 datasets related to Parkinson's disease from the UCI
machine learning library in identifying the Parkinson disease. In the specified dataset, there were 24 attributes. After training the
data, they were able to identify the machine learning algorithms that were most accurate. Naive Bayes, Adaptive Boosting, Bagging
Classifier, Decision Tree Classifier, Random Forest Classifier, XBG Classifier, K Nearest Neighbor Classifier, Support Vector
Classifier, and Gradient Boosting Classifier were the nine machine learning algorithms that were utilized to predict the illness.
Evaluation metrics analysis and confusion metrics analysis (Precision, Recall, F measure and Accuracy) have been used to calculate
the study's outcomes. Algorithm with highest accuracy was found using the above- mentioned metrics analysis.
Mohesh T et al.[3] The input is the Parkinson's disease voice dataset from the UCI device mastering library. Additionally, by
combining the spiral drawing inputs of healthy individuals and Parkinson's patients, the gadget delivers accurate findings. It can be
inferred that a hybrid approach accurately reads affected individuals' spiral drawings and voice data. This model aims to make this
method of expertise a case of Parkinson’s hence, the goal is to apply numerous machines getting to know strategies like SVM,
choice Tree, for buying the maximum accurate result.

©IJRASET: All Rights are Reserved | SJ Impact Factor 7.538 | ISRA Journal Impact Factor 7.894 | 696
 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.538
Volume 11 Issue III Mar 2023- Available at www.ijraset.com

Ifeoma Oduntan[4] implemented XGBoost (Extreme Gradient Boosted Algorithm) based on its accurate application . Precision and
recall metrics were used in evaluating the performance of the classifier . Python 3.8 was used in the implementation of this project
because of its flexibility, huge collection of libraries and it is an open-source language. The XGBoost algorithm was applied using
different parameter tunings to find the best estimator. GridSearchCV was used, and scoring was set to f1. The best score and
parameters were extracted, and this was used in fitting the model. Confusion matrix was used in evaluating the performance of the
classifier.
Sonia Singla (2021)[5] Taken from the UCI repository. To determine the optimum algorithm for disease, start identification,
algorithms like XGBoost, KNN, SVMs, and Random Forest Algorithm were tested. Confusion matrix and accuracy score were used
to evaluate the models.
Anik Pramanik et al. (2020) [6] proposed a model to detect Parkinson’s Disease using voice and speech signal data in an efficient
and robust manner. The dataset, on which the proposed model was tested, was publicly available as PD Speech data-set from
Department of Neurology in Cerrahpa¸sa, Faculty of Medicine, Istanbul University. The dataset contained no less than 750
attributes of 252 persons. Data standardization, Multicollinearity Diagnosis, Principal Component Analysis, and Independent
Component Analysis were among the different data-processing methods used. Various algorithms used includes Support Vector
Machine, Logistic Regression and Random Forrest.
Anitha R at el. [7] suggested a predictive analytics system that uses K-means clustering and Decision Tree to extract insights from
patients. This specific study uses the Parkinson's disease speech dataset from the UCI Machine Learning library as its input. The
suggested approach also delivered precise outcomes by combining the spiral drawing inputs of Parkinson's patients and healthy
individuals. They suggested a hybrid methodology, which produced technique that detects after evaluating patient speech and spiral
drawing data. The drawings were converted into pixels using the Random Forest classification technique, and the extracted values
were compared to the training database to generate various characteristics.
Sanghee Moon et al.[8] This retrospective database study includes a total of 1468 people tested at the Parkinson’s Disease and
Movement Disorder Clinic of the University of Kansas Medical Center. A total of 130 balance and gait features were automatically
computed by the Mobility Lab software. Neural networks, Support Vector Machine, kNeareast Neighbors, decision trees, random
forests, gradient boosting classifiers, and linear regression were among the classification models. Accuracy, recall, precision, and F1
score were all used to evaluate the classification models.
Timothy J. Wroge et al. [9] Data were gathered through mPower, clinical observational research carried out by Sage Bionetworks
with the use of an iPhone app. Before being fed into the feature extraction algorithms, the raw audio is cleaned with VoiceBox's
Voice Activation Detection (VAD) method. Scikit-Learn was used to create the decision tree and support vector machine classifiers.
Several decision tree classifiers, including additional trees, random forests, gradient boosted decision trees, and normal decision
trees, were utilized to categorize the dataset.
Basil K Varghese at el. [10] used the UCI ML repository to access the dataset. The dimension of the data was then reduced using
Principal Component Analysis. They used SVM (Support Vector Machine), Decision Trees, Linear Regression, and Neural
Networks to predict values from the test dataset. The accuracy was then determined by using the training model to predict values
from the test dataset. The dataset used in this study consists of characteristics extracted from voice recordings of 42 individuals who
have been diagnosed with early-stage Parkinson's disease. The goal of this work was to use a variety of machine learning techniques
to reliably predict the RMSE (Root Mean Square Error) values of motor and total UPDRS scores (referred to as "motor UPDRS"
and "total UPDRS").
Srishti Grover at al. [11] Proposed a way for applying deep learning to forecast the severity of Parkinson's illness. In the first stage,
voice recordings from PD patients are obtained for analysis. The obtained data is then normalized using min-max normalization.
The model after getting data performs training, evaluation, and prediction. A deep neural network containing an input layer, hidden
layers, and an output layer is created. And in the end, an evaluation is performed on the resultant DNN classifier.
Tarigoppula V. S Sriram at el. [12] proposed research on Diagnosis of the Parkinson disease through machine learning approaches.
Orange v2.0b and weka v3.4.10 was used in the experiment for the statistical analysis, classification, Evaluation, and unsupervised
learning methods. A voice dataset for Parkinson's disease was collected from the Center of Machine Learning and Intelligent
Systems at UC Irvine. Of the 31 persons in the collection who had biological voice measurements, 23 of them had the disease. The
dataset was acquired and utilized for data visualization (parallel coordinates, Sieve graphs, and SOM), classification (Bayes Net,
Nave Bayes, Logistic, Simple Logistic, K Star, AD Tree, J48, LMT, and Random Forest), as well as for evaluation and
unsupervised learning techniques (Hierarchal clustering).

©IJRASET: All Rights are Reserved | SJ Impact Factor 7.538 | ISRA Journal Impact Factor 7.894 | 697
 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.538
Volume 11 Issue III Mar 2023- Available at www.ijraset.com

II. COMPARISON AMONG MODELS

We compared the work based on metrics, algorithms used and the accuracy on several datasets used by the authors. The work is
summarized as shown in Table 1.
Table-1: Comparison Among Models
Reference Dataset Tools used Machine Learning Evaluation Metrics Outcomes of Outperformed
Algorithms Used Algorithms

Amreen Khanum D UCI Machine Learning Decision Tree Classifier, Recall, Precision, F1- KNN reached to the highest
et al. [1] Repository Logistic Regression , Naive Score , R2-Score and accuracy:
2022 --- Bayes, KNN Classifier and accuracy Accuracy=0.966102 F1-
XGBoost Classifier Score=0.960000 Recall=
0.923077 Precision=1.00000
R2-Score=0.862471
Mustasim Shafi 195 Parkinson's disease Python and the Naive Bayes, Adaptive Confusion metrics A]Confusion metrics analysis:
Kader datasets took from the UCI Scikit-learn Boosting, Bagging analysis and K Nearest Neighbor with 97%
et al. [2] machine learning module. Classifier, Decision Tree Evaluation metrics accuracy. Predictive
2022 repository. Classifier, Random Forest analysis using Positive=6,0 Predictive
classifier, XBG Classifier, Precision, Recall , F Negative=1,32
K Nearest Neighbor measure and B]Evaluation metrics Analysis :
Classifier, Support Vector Accuracy K Nearest Neighbor with 97%
Classifier and Gradient accuracy Precision=1.00
Boosting Classifier Recall=0.86 F1-
Score=0.92 TP=1.00
TN=0.97
Mohesh T UCI Parkinson Dataset. Logistic Regression, PCA and EDA Decision Tree with an accuracy
et al. [3] Python 3 Support Vector Machine (TSNE visualization) of 94.7774555%.
2022 (SVM) , Ada Boost , with accuracy
Gradient Boost, Random
Forest, Naive Bayes, Neural
Network , XGBoost and
Decision Tree
Ifeoma Oduntan et Vocal measurement of 195 XGBoost Accuracy, Precision Accuracy=95% Precision=1.00
al. [4] instances and 24 attributes Python 3.8 and Recall Recall=0.94
2021 from 31 people and 23 of
them have Parkinson’s
Disease.
Sonia Singla [5] UCI Parkinson Dataset Python 3 XGBoost, KNN, SVMs, Confusion matrix and XGBoost with an accuracy of
2021 and Random Forest accuracy score 94%
Algorithm Matrix:
True healthy:[2 5]
True Parkinsons:[0 32]
Anik Pramanik et PD Speech data-set from SVM ,Logistic Regression PCA and ICA for SMV showed the highest
al. [6] Department of Neurology ,KNN , Ada Boost and data overall performance with
2020 in Cerrahpa sa, Faculty of Random Forest Algorithm pre-processing and accuracy 94.1%.
Medicine, Istanbul ---- accuracy
University
Anitha R et al. [7] UCI Machine learning RStudio and K-Means ,Random Forest A]Voice Data A]Voice data Analysis:
2020 library Visual Studio and Decision Tree Analysis :Accuracy Accuracy of 88% (Clustering
Code B]Spiral Drawing and classification).
Analysis : Accuracy B]Spiral Drawing Analysis:
and Confusion matrix Accuracy of 83% (Random
Forest).
Sanghee Moon et Parkinson’s Disease and IMU sensors Neural network(NN), Accuracy, precision , The F1-score of NN was
al. [8] Movement Disorder Clinic Support vector machines , recall and F1-score. 0.61,Precision of 0.61 ,Recall of
2020 of the University of Kansas k-nearest neighbor , 0.61 and accuracy of 89%
Medical Center. Decision tree , Random ,showed the highest
forest , performance among 8 models.
Gradient boosting and
Logistic regression

©IJRASET: All Rights are Reserved | SJ Impact Factor 7.538 | ISRA Journal Impact Factor 7.894 | 698
 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.538
Volume 11 Issue III Mar 2023- Available at www.ijraset.com

Timothy J. Wroge The data used for this Scikit-Learn Decision tree ,Support Cross validation with Gradient Boosted Decision Tree
et al. [9] analysis were collected machine vector machine ,Extra Trees accuracy, F-1, recall with 86% and 82% accuracy ,F-
2019 through mPower, a clinical learning library ,Gradient Boosted Decision and precision 1 score of 0.79 and
observational study as well as the Tree ,Artificial Neural 0.71,Precision of 0.85 and
conducted by Sage TensorFlow Network and Random 0.789, recall score of 0.73 and
Bionetworks using an and Keras Forest. 0.65 for AVEC and GeMaps
iPhone app Deep Learning features.
Libraries
Basil K Varghese Parkinson’s Support Vector Regression , Root Mean Squared Support Vector Regression
et al. [10] Telemonitoring dataset ---- Decision Tree Regression Error (RMSE) of demonstrated the best results
2019 from UCI ML repository ,Linear Regression and Motor and total with least RMSE values :
Resilient Back Propogation UPDRS scores , 7.49(Total UPDRS) and
6.06(Motor UPDRS)
Srishti Grover Parkinson’s The Python Deep neural networks Motor-UPDRS and 62.7335% accuracy with total
et al. [11] Telemonitoring Voice Data library, Total-UPDRS UPDRS score and 81.6667%
2018 Set from UCI Machine TensorFlow accuracy accuracy with motor UPDRS
Learning Repository (tf.estimator) score
Tarigoppula V.S UCI Machine learning Orange Bayes Net , Naïve Bayes , Parallel coordinates , Random Forest with an
Sriram et al. [12] repository from Centre for software v2.0b Logistic ,Simple Logistic , Sieve graph ,ROC accuracy of 90.26%
2013 Machine Learning and and weka KStar , ADTree, visualization and outperformed other algorithms
Intelligent Systems v3.4.10 J48, LMT and accuracy
Random Forest

III. CONCLUSION
Managing Parkinson disease in day-to-day life is very challenging for an individual. A good screening procedure will be beneficial,
especially in circumstances where a physician's treatment is not necessary. We discovered several researchers engaged in the field of
Parkinson disease detection during the survey. A decision to select a specific system from the pool of available researchers can be
made based on the requirements and resources available.

REFERENCES
[1] Amreen Khanum D, Prof. Kavitha G and Prof. Mamatha H S, “Parkinson’s Detection using Machine Learning Algorithms”, International Journal for
Research in Applied Science & Engineering Technology (IJRASET), ISSN: 2321- 9653 vol 10 Issue X, pp 786-790, Oct 2022.
[2] Muhtasim Shafi and Fizar Ahmed, “Parkinson’s Disease Detection Analysis through Machine Learning Approaches”
https://www.researchgate.net/publication/359711136, 2022.
[3] Mohesh T,Gowtham K,Vijeesh P and Arun Kumar S, “Parkinson’s Disease Prediction Using Machine Learning”,IJRASET44075,ISSN:2321-9653,vol 10 Issue
VI,2022.
[4] Oduntan Ifeoma, “Prediction of Parkinson's Disease Using Biomedical Voice Measurements Dataset”, https://www.researc
hgate.net/publication/35725620,2021.
[5] Sonia Singla, “Parkinson disease onset detection Using Machine Learning”, https://www.analyticsvidhya.com/blog/2021/07/parkinson-disease-onset
detection-using-machine-learning/,2022.
[6] Pramanik Anik and Sarker Amlan, “Parkinson's Disease Detection from Voice and Speech Data Using Machine Learning”,
https://www.researchgate.net/publication/347520593,2020.
[7] Anitha R,Nandhini T, Sathish Raj S and Nikitha V, “Early Detection Of Parkinson’s Disease Using Machine Learning”,IJARIIE,ISSN(O):2395-4396, Vol 6
Issue 2, pp 505-511,2020.
[8] Sanghee Moon, Hyun-Je Song, Vibhash D. Sharma, Kelly E. Lyons, Rajesh Pahwa, Abiodun E. Akinwuntan and Hannes Devos. “Classification of Parkinson’s
disease and essential tremor based on balance and gait characteristics from wearable motion sensors via machine learning techniques: a data-driven approach”
Journal of Neuro Engineering and Rehabilitation, ISSN: 1743-0003, vol 17, Issue 1,2020.
[9] T. J. Wroge, Y. Özkanca, C. Demiroglu, D. Si, D. C. Atkins and R. H. Ghomi, “Parkinson’s Disease Diagnosis Using Machine Learning and Voice”,IEEE
Signal Processing in Medicine and Biology Symposium (SPMB), ISSN: 2372- 7241, pp. 1-7, 2018.
[10] Basil K Varghese, Geraldine Bessie Amali D and Uma Devi K S. “Prediction of Parkinson’s Disease using Machine Learning Techniques on
Speech dataset” Research Journal of Pharmacy and Technology, vol 12 Issue (2), pp 644- 648,2019.
[11] Srishti Grover, Saloni Bhartia, Akshama, Abhilasha Yadav and Seeja K. R, “Predicting Severity of Parkinson's Disease Using Deep Learning”, International
Conference on Computational Intelligence and Data Science (ICCIDS), vol 132, pp 1788–1794,2018.
[12] Tarigoppula V.S Sriram, M. Venkateswara Rao, G V Satya Narayana, DSVGK Kaladhar and T Pandu Ranga Vital, “Intelligent Parkinson Disease
Prediction Using Machine Learning Algorithms”, IJEIT, vol 3, pp 212-215,2013.

©IJRASET: All Rights are Reserved | SJ Impact Factor 7.538 | ISRA Journal Impact Factor 7.894 | 699