Computers, Materials & Continua

DOI:10.32604/cmc.2022.xxxxxx
Type: xxx

Enhancing Parkinson’s Disease Prediction Using Machine Learning and

Feature Selection Methods
Faisal Saeed1,2, *, Mohammad Al-Sarem1,3, Muhannad Al-Mohaimeed1, Abdelhamid Emara1,4,
Wadii Boulila1,5, Mohammed Alasli1 and Fahad Ghabban1
1College of Computer Science and Engineering, Taibah University, Medina 41477, Saudi Arabia
2
School of Computing and Digital Technology, Birmingham City University, Birmingham B4 7XG, United Kingdom
3
Information System Department, Saba’a Region University, Mareeb, Yemen
4
Computers and Systems Engineering Department, Al-Azhar University, Cairo 11884, Egypt
5
RIADI Laboratory, National School of Computer Sciences, University of Manouba, Manouba 2010, Tunisia
*
Corresponding Author: Faisal Saeed. Email: [email protected]
Received: XX Month 202X; Accepted: XX Month 202X

Abstract: Several millions of people suffer from Parkinson’s disease globally.

Parkinson’s affects about 1% of people over 60 and its symptoms increase with
age. The voice may be affected and patients experience abnormalities in speech
that might not be noticed by listeners, but which could be analyzed using recorded
speech signals. With the huge advancements of technology, the medical data has
increased dramatically, and therefore, there is a need to apply data mining and
machine learning methods to extract new knowledge from this data. Several
classification methods were used to analyze medical data sets and diagnostic
problems, such as Parkinson’s Disease (PD). In addition, to improve the
performance of classification, feature selection methods have been extensively
used in many fields. This paper aims to propose a comprehensive approach to
enhance the prediction of PD using several machine learning methods with
different feature selection methods such as filter-based and wrapper-based. The
dataset includes 240 recodes with 46 acoustic features extracted from 3 voice
recording replications for 80 patients. The experimental results showed
improvements when wrapper-based features selection method was used with K-
NN classifier with accuracy of 88.33%. The best obtained results were compared
with other studies and it was found that this study provides comparable and
superior results.

Keywords: Filter-based Feature Selection Methods; Machine Learning;

Parkinson’s disease; Wrapper-based Feature Selection Methods.

1 Introduction
Parkinson’s disease (PD) is a long term degenerative disorder of the central nervous system which causes
both motor and non-motor symptoms [1]. The exact causes of PD are unknown and unclear, but it is
supposed to include risk factors which are both genetic and environmental. More than 10% of patients with
PD have a first-degree relative with PD disease. In addition, PD is more prevalent between people who are
disclosed to some pesticides and the people with past history of head injury, while PD risk is lower for
patients who smoke [2]. PD mainly affects neurons in a certain region of the mid brain that is known as
substantia nigra, dopamine-producing brain cells, which leads to inadequate dopamine secretion in this
region [3].
In the early stage of the PD, the main symptoms are shaking, difficulty with walking and slowness of
movement. The common symptoms with late phase of PD are anxiety, dementia and depression. Moreover,
emotional problems, sleep and sensory symptoms may also occur [4, 5], in addition to Parkinsonian
syndrome [6]. These symptoms are mainly used to diagnose typical PD, in addition to examinations such

This work is licensed under a Creative Commons Attribution 4.0 International License, which
permits unrestricted use, distribution, and reproduction in any medium, provided the original work
  is properly cited.
xxxx CMC, 2022, vol.xx, no.xx

as neuroimaging. There is no total recovery for PD, however treatment aims to improve the symptoms [7,
8]. The medical decision support systems (MDSS) have increasingly used a significant diagnosis and
treatment method that uses artificial intelligence (AI) methods on a clinical dataset to assist clinicians to
make better decisions [9,10]. Recent improvements in machine learning, AI and statistical learning have
improved decision support system (DSS), which has helped to introduce intelligent decision systems
[10,11]. Some studies reported that the artificial intelligence cannot be effective without learning [12].
There are many types of machine learning methods such as Support Vector Machine (SVM), Naïve Bayes
(NB), K-nearest Neighbor (KNN), Multilayer Perceptron, Decision Tree (DT) and Random Forests (RF)
that have been used to solve medical decision problems.
There is a significant overlapping between ML and data mining which often use the same procedures,
but whereas ML concentrates on prediction, based on previously definite properties learned from the
training data, data mining concentrates on the detection of unknown properties in the clinical data. The
machine learning (ML) techniques have a significant role to play in the medical disease diagnosis field and
are widely used in bioinformatics [13, 14].
Recently, the variety of medical data is continuously increasing, therefore, effective classification and
prediction algorithms are required. The previous studies on machine learning research reported that the
accuracy of a classification algorithm can be influenced by many agents [15]. ML algorithms are used to
analyze medical data sets and diagnostic problems [12]. Subsequently, improvement of medical decisions,
treatments, and decrease financial costs will occur [16,14].
In addition, feature selection plays an important role in the explanation of medical data. Feature selection
technique constitutes a significant issue of global combinatorial optimization in machine learning, which is
used to decrease the number of features from the original features, removes irrelevant or redundant features
without incurring much loss of information, as well as simplification of models to make them easier to
interpret and shortening training times [17]. Therefore, a good feature selection method is required to
accelerate processing time and predictive accuracy. There are three types of feature selection algorithms,
which are: filter (extract features from the data set without any learning), wrapper (use learning techniques
to estimate useful features) and hybrid (gather the feature selection step and the classifier construction) [18,
19].
Recently, the medical field is the most favorable field to use machine learning methods. Therefore, Naïve
Bayes (NB), Support Vector Machine (SVM), K-nearest Neighbor, Multilayer Perceptron and Random
Forests as well as feature selection methods have been suggested to solve medical decision problems, such
as the prediction of Parkinson's disease. In this paper, the main contributions in the domain of prediction of
Parkinson’s Disease can be summarized as follows.
1.   A comprehensive approach was used to investigate the performance of several feature selection
methods and machine learning methods in order to enhance the prediction of PD.
2.   These feature selection methods include both filter-based methods such as (Information gain IG,
Principle Component Analysis PCA) and wrapper methods that include different search methods such
as First Best Greedy Stepwise PSO Method.
3.   A comparative analysis was conducted to examine the performances of all methods/combinations used
and the best prediction results were reported.
This paper is organized as follows: Section 2: the related works. Section 3: discussion of the methods.
Section 4: experimental results and discussion. Section 5: conclusions and future works.

2 Related Studies
Several works have investigated the diagnosis of PD, in which many machine learning methods were
applied such as Support Vector Machine, neural network, Naïve Bayes, K-nearest neighbor and Random
Forests. In this paper, several datasets were used to search for related studies on Parkinson’s disease,
including Scopus, IEEE Xplore, Science Direct and Google Scholar.
CMC, 2022, vol.xx, no.xx xxxx

In [20] a supervised ML method was proposed that combined the Principal Components Analysis
(PCA) to extract features and SVM as classification method to identify PD patients. The main goal of this
method was to determine patients that will be diagnosed with PD or with Progressive Supranuclear Palsy
(PSP). The experiments were conducted on data of several patients with clinical and demographic features.
The results depicted good accuracy of the proposed method in identifying the PD patients compared to
existing related works.
In addition, the authors in [21] proposed an expert system of PD using features extracted from
recordings of patients’ voice. They developed a Bayesian classification approach to deal with the
dependence to match the replication-based experimental design. The experiments were performed on voice
recordings involving 80 subjects, 50% of them had PD. The aim was to identify which subjects had no PD
and which did have the disease. Naranjo et al. addressed the problem of identifying PD patients using the
extracted acoustic features from repeated voice recordings. The proposed method was based on two steps,
namely variable selection, and classification. The first step aims to reduce the number of features, while the
next step uses a regularization method named LASSO (Least Absolute Shrinkage and Selection Operator)
as a classifier. The proposed method was tested on the previously described database and showed a good
capacity for PD discrimination.
In addition, the authors in [22] addressed the problem of PD diagnosis by developing an approach that
investigated gait and tremor features that were extracted from the voice reordering data. They started by
filtering data to remove noises, then, using this data to extract gait features they detected the peak and
measured the pulse duration. The average accuracy obtained for the identifying PD patients by the proposed
approach was satisfactory.
The authors in [23] proposed a method to automatically detect PD by using the convolutional neural
network (CNN). The authors suggested considering electroencephalogram (EEG) signals to build a thirteen-
layer CNN model. The proposed approach experimented with EEG signals of 20 Parkinson’s disease
patients (50% men and 50% women). The CNN method obtained interesting results to identify PD patients;
however, its performance should be evaluated using a large population.
Recently, Mostafa et al. [24] tried to enhance the diagnoses of PD by using several methods of feature
evaluation and classification. They used a multi-agent system to evaluate multiple features by using five
classification methods, namely DT, NB, NN, RF, and SVM. To evaluate the proposed method, they
conducted several experiments using original and filtered datasets. The results depicted that this method
enhanced the performance of ML methods used by finding the best set of features.
In addition, several methods were applied by [25-27] in order to predict Parkinson’s disease]. These
methods applied several machine learning and feature selection methods to enhance the prediction of
Parkinson’s disease and other studies utilized machine learning and deep learning to improving prediction
of diseases [28-38]. This paper extends these efforts by applying a comprehensive approach to investigate
the performance of several machine learning with feature selection methods.

3 Methods
There are many feature selection techniques available, and we have considered the utilization of the
following feature selection techniques: Filter-based technique, Correlation-based Feature Subset Selection
(CfsSubsetEval), Principle Component Analysis (PCA), and Wrapper technique. The aforementioned
techniques use different strategies or search algorithms to generate subsets and progress the search
processes including (i) Best First (ii) Greedy Stepwise, (iii) Particle Swarm Optimization (PSO), and (vi)
Ranker (see Fig. 1).
The dataset used in this paper is available online at UCI Machine Learning Repository [14]. The
dataset contains acoustic features of 80 patients, 50% of them suffering from Parkinson’s disease. The data
set has 240 recordings with 46 acoustic features extracted from 3 voice recording replications per patient.
The data set is well-balanced by gender and class label (whether the patients have Parkinson’s disease or
not).
xxxx CMC, 2022, vol.xx, no.xx

The experimental protocol was designed for evaluating the combination of the above techniques and
search algorithms when they were used with the following classification models: (i) Naïve Bayes, (ii)
Support Vector Machine (SVM)1, (iii) K-Nearest Neighbor (K-NN), (vi)Multi-Layer Perceptron (MLP)
and (v) Random Forest (RF). The experiments were carried out on WEKA tool version 3.8 and MacBook
Pro with OS X Yosemite version 10.10.5 as an operating system. To evaluate the performance of each
classifier, we first ran feature selection in order to find the representative features and then we applied the
classification models. Additionally, 10-fold cross validation was applied and the results have been reported
in terms of Accuracy, Recall, Precision and F-score. Finally, we analyzed the results achieved from the
experimentations. As stated earlier, the main goal of the research is to enhance the prediction of Parkinson’s
disease. However, this work also provides a useful guide to selecting the best feature selection technique
for different classification models.
   

Figure 1: Filter-based Approach vs. Wrapper-based Approach

3.1   Feature Selection Techniques

Several feature selection techniques were applied before feeding the data into the classifier. The filter-
based techniques consider the relevance between the features. Thus, they have low complexity, acceptable
stability and scalability [39]. A disadvantage of this type of technique is that it might ignore some
informative features, especially when the data is coming in stream [40]. The filter-based approaches can
be either univariate or multivariate [41]. The univariate methods examine features according to the
statistically-based criterion such as Information Gain (IG) [42-44]. Multivariate methods compute feature
dependency before ranking the feature. In addition, Principle Component Analysis (PCA) is a common
statistical method that is used for data analysis. PCA reduces the size of the data sets by selecting a set of
features that represents the whole data set. Since PCA is a conversion technique, the principal components
of the first variables is the component with the highest variance value. Then, other principal components
are ordered with descending variance values [45]. In addition, the wrapper-based techniques evaluate the
quality of the selected features using the performance of the learning classifier.

1
Both, the c-SVM and nu-SVM are examined.
CMC, 2022, vol.xx, no.xx xxxx

Regarding the search strategies, the search algorithms follows either sequential forward search (SFS),
or sequential backward search (SBS). The SFS starts with a single feature and then iteratively adds or
removes features until some terminating criterion is met whereas SBS starts with the whole feature set and
then continues with adding and deleting operations. Since the SBS method attempts to find solutions ranged
between suboptimal and near optimal regions [41], it is worth fully employing optimization techniques to
figure out the subset that leads to maximizing the learner’s performance, in particular, with the wrapper
approach. At this end, the wrapper-based method can take advantage of various optimization methods such
genetic algorithm [46,47] and ant colony optimization algorithm (ACO) [48].

3.2   Machine Learning Classifiers

In machine learning, the data classification is still an attractive domain. Lately, there are many
proposed algorithms that have been examined in several domains such as NB, SVM, K-NN, MLP and RF,
which are presented briefly in the next subsections.

3.2.1  Support  Vector  Machine    

The basic idea behind SVM algorithm is to construct a hyperplane between groups of data. The quality
of the hyperplane is evaluated by measuring to which degree it can maintain the largest distance from the
points in either class [39]. Therefore, as it is presented in Fig.2, the higher the separation ability of the
hyperplane, the lower the error in the value [49]. The computational complexity of SVM is 𝑂(𝑛$  ) [50,51].

Figure 2: SVM illustration. The larger margin separating the data points, the higher accuracy we
obtained.

3.2.2  Naïve  Bayes  

Naïve Bayes (NB) is a probabilistic classifier that is based on Bayesian theorem. It is called Naïve
because the classifier works on a strong features independence assumption. In literature, there are several
variants of NB: simple Naïve Bayes, Gaussian Naïve Bayes, Multinomial Naïve Bayes, Bernoulli Naïve
Bayes and Multi-variant Poisson Naïve Bayes in which the main different among them is the way the
probability of the target class is computed. The time complexity of Naïve Bayes 𝑖𝑠  𝑂 (𝑑×𝑐) where 𝑑 is the
query vector's dimension, and 𝑐 is the total classes.

3.2.3  K-­‐‑Nearest  Neighbor    

xxxx CMC, 2022, vol.xx, no.xx

K-NN is a type of lazy learning, in which there is no explicit training phase and all computations are
deferred until classification. It is a method of classifying data based on the nearest training data points in
the feature space. The K-NN classifier uses the Euclidean distance measure, or another measure such as
Euclidean squared, Manhattan, and Chebyshev, to estimate the target class. The performance of the
classifier depends upon the parameter k, while the best value of k depends upon the dataset. In general, the
greater the value of k, the lower the noises in the classification, but the boundaries between the classes
become less distinct as shown in Fig. 3. The time complexity of K-NN is 𝑂(𝑛×𝑚), where n is the number
of training examples and m is the number of dimensions in the training set [52].

Figure 3: K-NN Model. When k=3, the classifier predicts a new point as B class (Fig. a), whilst, when
k=5, the point is determined as a class A.

3.2.4  Multilayer  Perceptron  Model  

The MLP is a classical feedforward neural network classifier in which the errors of the output are used
to train the network [53]. MLP consists of three layers of nodes: (i) input layer, (ii) at least one or more
hidden layer(s), and (iii) output layer. The input layer is connected to the hidden layers which are connected
to the output layer. All the layers are processed by weighted values. Fig. 4 represents a MLP with a single
hidden layer. MLP is one-way error propagation where back-propagation techniques have been utilized to
train and test these weight values. The time complexity of MLP is 𝑂(𝑛$ ).

Figure 4: MLP Model with 1 input layer, 1 hidden layer, and 1 output layer.

3.2.5  Random  Forests  

CMC, 2022, vol.xx, no.xx xxxx

The Random Forests (RF) classifier is a type of ensemble method that combines multiple decision tree
predictions. In RF, the trees are generated randomly by selecting attributes at each node. The output of the
ensemble is tree votes with the most popular class. The pseudo-code of the Random Forest ensemble is
presented in Tab. 1. The time complexity of Random Forest of size 𝑇 and maximum depth 𝐷 (excluding
the root) is 𝑂(𝑇×𝐷) [54].
Table 1: Pseudo-code of RF model.
Input:
𝐷 / , a set of 𝑑 training sets;
𝐷 // , a set of 𝑑 test sets;
𝑘, number of models in the ensemble;
𝑓, number of attributes that are used to split the 𝐷 /
𝑀3 , set of base classifiers.
Output:
The ensemble (a composite model 𝑀 ∗ )
Steps:
(1) for 𝑖 = 1 to 𝑘 do
(2) - create bootstrap sample of 𝐷 / with replacement, 𝐷3 ;
(3) - construct a decision tree classifier by selecting randomly the attributes 𝑓;
(4) - use CART method to grow the trees.
(5) end for
//To use the ensemble for classifying data on test set:
(1) each of the k models (decision tree) classifies 𝐷 // and return the majority vote
 
The random forest method is more robust to errors and outliers. Therefore, the problem of over-fitting
is not faced. The accuracy of the model depends mainly on the strength of the base classifiers and measure
of the dependence between them [55].

4 Experimental Results
The experiments were conducted such that 10-fold cross validation was applied for each classifier.
The performance of each classifier was measured by the accuracy, precision, recall and F-score. Tab. 2 to
12 show the experimental results of several machine learning methods both with and without different
feature selection methods.

Table 2: The performance of classifiers without features selection

Performance →
Accuracy Precision Recall F-score
Classifier↓
Naïve Bayes 0.829 0.830 0.829 0.829
Support c-SVM 0.779 0.781 0.779 0.779
Vector nu-SVM 0.779 0.779 0.779 0.779
Machine
Multilayer Perceptron 0.767 0.767 0.767 0.767
K-nearest Neighbor 0.800 80.00 80.00 80.90
Random Forest 0.800 80.00 80.00 80.00
xxxx CMC, 2022, vol.xx, no.xx

Table 3: Performance of classifiers with CfsSubsetEval Feature Selection Combinations

Search
Classifier Accuracy Precision Recall F-score
Method
NB 0.817 0.817 0.817 0.817
c-SVM 0.767 0.767 0.767 0.767
Nu-SVM 0.762 0.763 0.763 0.762
First Best

MLP 0.758 0.759 0.758 0.758

K-NN 0.733 0.733 0.733 0.733
RF 0.804 0.805 0.804 0.804
NB 0.825 0.825 0.825 0.825
Greedy Stepwise

c-SVM 0.746 0.746 0.746 0.746

Nu-SVM 0.754 0.754 0.754 0.754
MLP 0.746 0.746 0.746 0.746
K-NN 0.742 0.743 0.742 0.741
RF 0.829 0.832 0.829 0.829
NB 0.821 0.821 0.821 0.821
c-SVM 0.754 0.754 0.754 0.754
Nu-SVM 0.746 0.746 0.746 0.746
PSO Method

MLP 0.738 0.738 0.738 0.737

K-NN 0.700 0.701 0.700 0.700
RF 0.821 0.822 0.821 0.821

Figure 5: Number of remaining features after applying features selection methods

Table 4: Performance of classifiers with features selection based on information gain

CMC, 2022, vol.xx, no.xx xxxx

Classifier Accuracy Precision Recall F-score

NB 0.808 0.808 0.808 0.808
c-SVM 0.696 0.697 0.696 0.695
Nu-SVM 0.700 0.701 0.700 0.700
MLP 0.738 0.738 0.738 0.737
K-NN 0.708 0.712 0.708 0.707
RF 0.800 0.801 0.800 0.800

Table 5: Performance of classifier with features selection based on PCA

Classifier Accuracy Precision Recall F-score
NB 0.733 0.734 0.733 0.733
c-SVM 0.833 0.834 0.833 0.833
Nu-SVM 0.838 0.838 0.838 0.837
MLP 0.746 0.746 0.746 0.746
K-NN 0.692 0.696 0.692 0.690
RF 0.804 0.807 0.804 0.804

Table 6: Summary of the accuracy of classifiers with filter-based features selection methods
Best
Accuracy
Classifier Accuracy Method
(before FS)
with FS
NB 82.92 82.92 IG
c-SVM 77.92 83.33 PCA
nu-SVM 77.92 83.75 PCA
MLP 76.67 76.67 IG
K-nn 80.00 80.00 IG
RF 80.00 80.42 CfsSubsetEval With Greedy
Stepwise

Table 7: Performance of classifiers for wrapper-based method with Naïve Bayes as base classifier
Search
Classifier Accuracy Precision Recall F-score
Method
NB 0.838 0.838 0.838 0.837
First Best / Greedy

c-SVM 0.721 0.721 0.721 0.721

Nu-SVM 0.746 0.746 0.746 0.746
MLP 0.800 0.800 0.800 0.800
Stepwise

K-nn 0.758 0.758 0.758 0.758

RF 0.792 0.792 0.792 0.792
NB 0.854 0.855 0.854 0.854
Method

c-SVM 0.717 0.718 0.717 0.716

PSO

Nu-SVM 0.721 0.721 0.721 0.721

xxxx CMC, 2022, vol.xx, no.xx

MLP 0.761 0.763 0.763 0.762

K-nn 0.792 0.792 0.792 0.792
RF 0.738 0.738 0.738 0.737

Table 8: Performance of classifiers for wrapper-based methods with c-SVM as base classifier
Search
Classifier Accuracy Precision Recall F-score
Method
NB 0.833 0.834 0.833 0.833
c-SVM 0.850 0.850 0.850 0.850
Nu-SVM 0.846 0.846 0.846 0.846
First Best

MLP 0.788 0.788 0.788 0.787

K-NN 0.775 0.775 0.775 0.775
RF 0.817 0.820 0.817 0.816
NB 0.800 0.803 0.800 0.799
Greedy Stepwise

c-SVM 0.850 0.851 0.850 0.850

Nu-SVM 0.850 0.851 0.850 0.850
MLP 0.838 0.839 0.838 0.837
K-NN 0.775 0.783 0.775 0.773
RF 0.813 0.814 0.813 0.812
NB 0.8212 0.824 0.821 0.82
c-SVM 0.842 0.842 0.842 0.842
Nu-SVM 0.846 0.846 0.846 0.846
PSO Method

MLP 0.808 0.810 0.808 0.808

K-NN 0.783 0.784 0.783 0.783
RF 0.821 0.823 0.821 0.821

Table 9: Performance of classifiers for wrapper-based methods with nu-SVM as base classifier
Search
Classifier Accuracy Precision Recall F-score
Method
NB 0.754 0.778 0.754 0.749
c-SVM 0.779 0.786 0.779 0.778
Nu-SVM 0.779 0.782 0.779 0.779
First Best

MLP 0.808 0.813 0.808 0.808

K-NN 0.775 0.775 0.775 0.775
RF 0.779 0.780 0.779 0.779
NB 0.754 0.778 0.754 0.749
c-SVM 0.779 0.786 0.779 0.778
Nu-SVM 0.779 0.782 0.779 0.779
Stepwise
Greedy

MLP 0.808 0.813 0.808 0.808

K-NN 0.775 0.775 0.775 0.775
CMC, 2022, vol.xx, no.xx xxxx

RF 0.779 0.780 0.779 0.779

NB 0.813 0.813 0.813 0.812
c-SVM 0.817 0.817 0.817 0.817
Nu-SVM 0.796 0.796 0.796 0.796

PSO Method
MLP 0.746 0.746 0.746 0.746
K-NN 0.817 0.818 0.817 0.816
RF 0.813 0.813 0.813 0.812

Table 10: Performance of classifiers for wrapper-based methods with MLP as base classifier
Search
Classifier Accuracy Precision Recall F-score
Method
NB 0.813 0.819 0.813 0.812
c-SVM 0.733 0.741 0.733 0.731
Nu-SVM 0.792 0.800 0.792 0.790
First Best

MLP 0.829 0.832 0.829 0.829

K-NN 0.771 0.771 0.771 0.771
RF 0.825 0.825 0.825 0.825
NB 0.813 0.814 0.813 0.812
Greedy Stepwise

c-SVM 0.733 0.739 0.733 0.732

Nu-SVM 0.800 0.809 0.800 0.799
MLP 0.825 0.826 0.825 0.825
K-NN 0.792 0.739 0.792 0.791
RF 0.821 0.824 0.821 0.820
NB 0.800 0.804 0.800 0.799
c-SVM 0.754 0.756 0.754 0.754
Nu-SVM 0.771 0.771 0.771 0.771
PSO Method

MLP 0.829 0.830 0.829 0.829

K-NN 0.800 0.800 0.800 0.800
RF 0.817 0.817 0.817 0.817

Table 11: Performance of classifiers when wrapper-based methods with K-NN are applied
Search
Classifier Accuracy Precision Recall F-score
Method
NB 0.763 0.781 0.763 0.758
c-SVM 0.742 0.746 0.742 0.741
Nu-SVM 0.771 0.773 0.771 0.77
First Best

MLP 0.792 0.792 0.792 0.792

K-NN 0.883 0.883 0.883 0.883
RF 0.821 0.825 0.821 0.82
xxxx CMC, 2022, vol.xx, no.xx

NB 0.788 0.801 0.788 0.785

Greedy Stepwise
c-SVM 0.704 0.706 0.704 0.704
Nu-SVM 0.679 0.68 0.679 0.679
MLP 0.783 0.786 0.783 0.783
K-NN 0.846 0.848 0.846 0.846
RF 0.788 0.790 0.788 0.787
NB 0.775 0.793 0.775 0.771
c-SVM 0.754 0.756 0.754 0.754
Nu-SVM 0.758 0.759 0.758 0.758
PSO Method

MLP 0.850 0.850 0.850 0.850

K-NN 0.883 0.884 0.883 0.883
RF 0.804 0.806 0.804 0.804

Table 12: Performance of classifiers for wrapper-based methods with RF as base classifier
Search
Classifier Accuracy Precision Recall F-score
Method
NB 0.775 0.776 0.775 0.775
c-SVM 0.721 0.723 0.721 0.720
Nu-SVM 0.767 0.767 0.767 0.767
First Best

MLP 0.796 0.796 0.796 0.796

K-nn 0.788 0.778 0.788 0.787
RF 0.838 0.839 0.838 0.837
NB 0.800 0.800 0.800 0.800
Greedy Stepwise

c-SVM 0.721 0.724 0.721 0.720

Nu-SVM 0.729 0.731 0.729 0.729
MLP 0.779 0.780 0.779 0.779
K-nn 0.779 0.731 0.729 0.729
RF 0.808 0.809 0.808 0.808
NB 0.817 0.818 0.817 0.816
c-SVM 0.742 0.743 0.742 0.741
Nu-SVM 0.763 0.763 0.763 0.762
PSO Method

MLP 0.838 0.838 0.838 0.837

K-nn 0.800 0.800 0.800 0.800
RF 0.838 0.839 0.838 0.837

Discussion
Tab. 2, shows the performance of all classifiers used before applying features selecting methods. The
results showed Naïve Bayes obtained the best performance using all evaluation measures compared to the
other classifiers. It obtained 82.92%, 83.30%, 82.90% and 82.90 % for accuracy, precision, recall and F-
score respectively.
CMC, 2022, vol.xx, no.xx xxxx

The number of features was reduced using correlation based feature selection (CfsSubsetEval) method
to 23, 17, 18 for the search methods of First Best, Greedy Stepwise and POS respectively, as shown in Fig.
5. The performance of with CfsSubsetEval combinations for each classifier is shown in Tab. 3. The results
showed that no improvements were obtained by most of the combinations, except for RF with Greedy
Stepwise and POS methods.
Tab. 4 showed the performance of classifiers used when features selection method based on
information gain was applied. As shown in Fig. 5, the number of features was reduced to 10. The results
showed that no improvements were reported on the performance of all classifiers after applying this feature
selection method.
In addition, Tab. 5 shows the performance of all classifiers when features selection method based on
PCA was applied. The results showed that only SVM methods obtained better performance after applying
this features selection method. The number of features was reduced to 20 as shown in Fig. 5.
Tab. 6 summarizes the performance of filter based features selection methods. The results showed that
feature selections with PCA obtained the best performance when SVM classifier was applied.
Tab. 7-12 show the performance of wrapper-based features selection methods using different base
classifiers. In each table, First Best, Greedy Stepwise and PSO search methods were applied.
Tab. 7 showed that, when Naïve Bayes was used as the base classifier for wrapper-based feature
selection method, the performance of NB using PSO search method was enhanced to 0.854, 0.855, 0.854
and 0.854 for accuracy, precision, recall and F-score respectively. The performance of the other classifiers
using this method was reduced.
Tab. 8 shows the performance of classifiers when the wrapper-based features selection method with
c-SVM as the base classifier was applied. The results showed the enhancements obtained by all classifiers
using all search methods. However, the best performance was obtained by SVM using First Best and Greedy
Stepwise search methods.
However, Tab. 9 shows the performance of classifiers when wrapper-based features selection method
with nu-SVM as the base classifier was applied. The results showed that the enhancements were obtained
by applying c-SVM, K-NN and RF, especially when the POS search method was used.
In addition, Tab. 10 shows the performance of classifiers when wrapper-based features selection
method with MLP as base classifier was applied. The results showed that the enhancements were obtained
by applying MLP and RF for the three search methods. The best results were obtained using MLP classifier.
Moreover, Tab. 11 shows the performance of classifiers when wrapper-based features selection
method with K-NN as base classifier was applied. The results showed that the enhancements were obtained
by applying K-NN and RF for the First Best and POS search methods. The best results were obtained using
K-NN classifier with accuracy, precision, recall and F-scores of 0.883, 0.884, 0.883 and 0.883 respectively.
Tab. 12 shows the performance of classifiers when wrapper-based features selection method with RF
as base classifier was applied. The results showed that the enhancements were obtained by applying MLP
and RF for the three search methods. The best results were obtained using RF classifier.
Tab. 13 shows a comparison of different wrapper-based features selection methods (using different
base classifiers). The results showed that the best performing classifier was K-NN associated with the
wrapper-based feature selection with KNN as base classifier, obtaining 88.33% accuracy. The number of
features was reduced (with the best performance obtained) to 20, 5 and 22 using First Best, Greedy Stepwise
and PSO search methods.
Table 13: Best Results for wrapper-based techniques
Accuracy Accuracy
Classifier Method
(before FS) with FS
NB 82.92 85.42 PSO
xxxx CMC, 2022, vol.xx, no.xx

c-SVS/nu-SVM with Greedy

c-SVM 77.92 85.00
Stepwise
nu-SVM 77.92 81.67 PSO with RF/MLP
MLP 76.67 82.92 Best First / PSO
K-NN 80.00 88.33 Best First / PSO
RF 80.00 83.75 Best First, PSO with MLP/RF

Finally, Tab. 14 shows a comparison of using different features selection methods (filter and wrapper
base methods). It shows that the best performance was obtained by K-NN classifier associated with
wrapper-based feature selection method with K-NN as base classifier and using Best First and PSO search
method.
Table 14: Comparison between Filter-based and Wrapper-based techniques
Filter- Wrapper-
Classifier Baseline Method Method
based based
NB 82.92 82.92 IG 85.42 PSO
c-SVS/nu-SVM with
c-SVM 77.92 83.33 PCA 85.00
Greedy Stepwise
nu-SVM 77.92 83.75 PCA 81.67 PSO with RF/MLP
MLP 76.67 76.67 IG 82.92 Best First / PSO
K-nn 80.00 80.00 IG 88.33 Best First / PSO
CfsSubsetEval With Best First, PSO with
RF 80.00 80.42 83.75
Greedy Stepwise MLP/RF

For this paper a comparison has been conducted between the best performing methods and the previous
studies on predicting Parkinson’s disease using the same dataset, and other datasets, as shown in Tab. 15.
The comparison results showed that the best performing method (K-NN classifier associated with wrapper-
based feature selection method with K-NN as base classifier and using Best First and PSO search method)
obtained comparable and superior results.

Table 15: Comparison with previous studies

Accuracy
Methods Dataset Used
with FS
The best
performing 88.33 UCI Dataset [14]
method
[21] 86.2 UCI Dataset [14]

[56] 57.5% Other Dataset

[57] 82.5% Other Dataset

[58] 82.5% Other Dataset
[59] 87.5% Other Dataset
CMC, 2022, vol.xx, no.xx xxxx

5 Conclusions and Future Works

This paper examined the performance of several classifiers with filter-based and wrapper-based
features selections methods to enhance the diagnosis of Parkinson’s disease. Different evaluation metrics
were used including accuracy, precision, recall and F-score. The experiments compared the performance of
machine learning on original and filtered datasets. The results showed that wrapper-based features selection
method with K-NN enhanced the performance of predicting Parkinson’s disease, with the accuracy reached
to 88.33%. In future work, more machine learning and deep learning methods could be applied with these
combinations of features selection methods. In addition, other features selection methods could be
investigated to improve the performance of predicting Parkinson’s disease.

Acknowledgement: The authors extend their appreciation to the Deputyship for Research & Innovation,
Ministry of Education in Saudi Arabia for funding this research work; project number (77 /442). Also, the
authors would like to extend their appreciation to Taibah University for its supervision support.

Funding Statement: This research was funded by the Deputyship for Research & Innovation, Ministry of
Education in Saudi Arabia under the project number (77 /442).

Conflicts of Interest: The authors declare that they have no conflicts of interest.

References
[1]   L. V. Kalia and A. E Lang, “Parkinson's disease,”. The Lancet, vol. 386, no. 9996, pp. 896–912, 2015.
[2]   J. P. Iannotti and R. Parker, “The netter collection of medical illustrations-musculoskeletal system,” in Elsevier
Health Sciences, 2nd edition, vol. 6, 2013.
[3]   M. Fjodorova, E.M. Torres and S.B. Dunnett, “Transplantation site influences the phenotypic differentiation of
dopamine neurons in ventral mesencephalic grafts in Parkinsonian rats," Experimental Neurology, vol. 291, pp.
8-19, 2017.
[4]   A. Jamak, A. Savatić and M. Can, “Principal component analysis for authorship attribution," Business Systems
Research, vol. 3, pp. 49-56, 2012.
[5]   M. Can, “Neural networks to diagnose the Parkinson ’s disease," Southeast Europe Journal of Soft
Computing,  vol.2, no. 1, 2013.
[6]   L.V. Kalia, S. K Kalia and A. E. Lang, "Disease modifying strategies for Parkinson's disease," Movement
Disorders, vol. 30, pp. 1442-1450, 2015.
[7]   N. Singh, V. Pillay and Y.E. Choonara, "Advances in the treatment of Parkinson's disease," Progress in
Neurobiology, vol. 81, pp. 29-44, 2007.
[8]   C. Camara, P. Isasi, K. Warwick, V. Ruiz, T. Aziz et al., “Resting tremor classification and detection in
Parkinson's disease patients," Biomedical Signal Processing and Control, vol. 16, pp. 88-97, 2015.
[9]   B. Keltch, L. Yuan and B. Coskun, “Comparison of AI techniques for prediction of liver fibrosis in hepatitis
patients,” Journal of Medical Systems, vol. 38, no. 8, pp. 1-8, 2014.
[10]   M. Nasr, K. El-Bahnasy, M. Hamdy and S.M. Kamal,   "A novel model based on non-invasive methods for
prediction of liver fibrosis," In 13th International Computer Engineering Conference (ICENCO), Cairo, Egypt,
27-28 Dec. 2017.
[11]   S. Guerlain, D.E. Brown and C. Mastrangelo, “Intelligent decision support systems”, in Proceeding of SMC 2000
conference proceedings. 2000 IEEE International Conference on Systems, Man and Cybernetics. 'cybernetics
evolving to systems, humans, organizations, and their complex interactions'. IEEE International Conference on
Systems, Man, and Cybernetics, Nashville, TN, USA, 2000, pp. 193438.
[12]   F. Meherwar and M. Pasha, “Survey of machine learning algorithms for disease diagnostic,” Journal of
Intelligent Learning Systems and Applications, vol. 9, no.1, 2017.
[13]   S. Brunak, “The Bioinformatics: Machine Learning Approach”, MIT press, 2001.
[14]   L. Naranjo, C.J. Perez, Y. Campos-Roca and J. Martin, “Addressing voice recording replications for Parkinson’s
disease detection," Expert Systems with Applications, vol. 46, pp. 286-292, 2016.
[15]   B. Harish, D. C. Hoyle and S. Singh, “Machine learning in bioinformatics: a brief survey and recommendations
xxxx CMC, 2022, vol.xx, no.xx

for practitioners,” Computers in Biology and Medicine, vol. 36, no.10, pp. 1104-1125, 2006.
[16]   R. Fernandez-Millan, J. A. Medina-Merodio, R. B. Plata, J. J. Martinez-Herraiz and J. M. Gutierrez-Martinez,
“A laboratory test expert system for clinical diagnosis support in primary health care,” Applied Sciences, vol.5,
no.3, pp. 222-240, 2015.
[17]   Y. Saeys, I. Inza and P. Larrañaga, “A review of feature selection techniques in bioinformatics,” Bioinformatics,
vol. 23, no.19, pp. 2507-2517, 2007.
[18]   Z. M. Hira and D. F. Gillies, “A review of feature selection and feature extraction methods applied on microarray
data,” Advances in Bioinformatics. Pp.198363, 2015.
[19]   P. Drotár and Z. Smékal, “Comparison of stability measures for feature selection,” in SAMI 2015IEEE 13th
International Symposium on Applied Machine Intelligence and Informatics, Herlany, Slovakia, 2015, pp.71-75.
[20]   C. Salvatore, A. Cerasa, I. Castiglioni, F. Gallivanone, A. Augimeri et al., “Machine learning on brain MRI data
for differential diagnosis of Parkinson's disease and progressive supranuclear palsy,” Journal of Neuroscience
Methods, vol. 222, pp.230-237, 2014.
[21]   L. Naranjo, C. J. Pérez, J. Martín and Y. Campos-Roca, “A two-stage variable selection and classification
approach for Parkinson’s disease detection by using voice recording replications,” Computer Methods and
Programs in Biomedicine, vol. 142, pp.147-156, 2017.
[22]   E. Abdulhay, N. Arunkumar, K. Narasimhan, E. Vellaiappan and V. Venkatraman, “Gait and tremor investigation
using machine learning techniques for the diagnosis of Parkinson disease,” Future Generation Computer Systems,
vol. 83, pp.366-373, 2018.
[23]   S. L. Oh, Y. Hagiwara, U. Raghavendra, R. Yuvaraj, N. Arunkumar et al., “A deep learning approach for
Parkinson’s disease diagnosis from EEG signals,” Neural Computing and Applications, vol. 32, no. 15, pp.
10927-10933, 2020.
[24]   S.A Mostafa, A. Mustapha, M.A. Mohammed, R.I. Hamed, N. Arunkumar et al., “Examining multiple feature
evaluation and classification methods for improving the diagnosis of Parkinson’s disease,” Cognitive Systems
Research, vol. 54, pp.90-99, 2019.
[25]   M. R. Salmanpour, M. Shamsaei, A. Saberi, S. Setayeshi, I. S. Klyuzhin et al., “Optimized machine learning
methods for prediction of cognitive outcome in Parkinson's disease,” Computers in Biology and Medicine, 111,
pp. 103347, 2019.
[26]   A. U. Haq, J. P. Li, M. H. Memon, A. Malik, T. Ahmad et al., “Feature selection based on L1-norm support vector
machine and effective recognition system for Parkinson’s disease using voice recordings,” IEEE Access, vol. 7,
pp.37718-37734, 2019.
[27]   C. Gao, H. Sun, T. Wang, M. Tang, N. I. Bohnen et al., “Model-based and model-free machine learning
techniques for diagnostic prediction and classification of clinical outcomes in Parkinson’s disease,” Scientific
Reports, vol. 8, no.1, pp.1-21, 2018.
[28]   H. Khalid, M. Hussain, M. A. A. Ghamdi, T. Khalid, K. Khalid et al., “A comparative systematic literature review
on knee bone reports from MRI, X-rays and CT scans using deep learning and machine learning
methodologies,” Diagnostics, vol. 10, no. 8, pp. 118–139, 2020.
[29]   M. A. Khan, S. Abbas, K. M. Khan, M. A. Ghamdi and A.Rehman, "Intelligent forecasting model of covid-19
novel coronavirus outbreak empowered with deep extreme learning machine," CMC-Computers, Materials &
Continua, vol. 64, no. 3, pp. 1329–1342, 2020.
[30]   A. H. Khan, M. A. Khan, S. Abbas, S. Y. Siddiqui, M. A. Saeed et al., “Simulation, modeling, and optimization
of intelligent kidney disease predication empowered with computational intelligence approaches,” Computers,
Materials & Continua, vol. 67, no. 2, pp. 1399–1412, 2021.
[31]   G. Ahmad, S. Alanazi, M. Alruwaili, F. Ahmad, M. A. Khan et al., “Intelligent ammunition detection and
classification system using convolutional neural network,” Computers, Materials & Continua, vol. 67, no. 2, pp.
2585–2600, 2021.
[32]   B. Shoaib, Y. Javed, M. A. Khan, F. Ahmad, M. Majeed et al., “Prediction of time series empowered with a novel
srekrls algorithm,” Computers, Materials & Continua, vol. 67, no. 2, pp. 1413–1427, 2021.
[33]   S. Aftab, S. Alanazi, M. Ahmad, M. A. Khan, A. Fatima et al., “Cloud-based diabetes decision support system
using machine learning fusion,” Computers, Materials & Continua, vol. 68, no. 1, pp. 1341–1357, 2021.
[34]   M. W. Nadeem, H. G. Goh, M. A. Khan, M. Hussain, M. F. Mushtaq et al., “Fusion-based machine learning
architecture for heart disease prediction,” Computers, Materials & Continua, vol. 67, no. 2, pp. 2481–2496, 2021.
[35]   S. Y. Siddiqui, I. Naseer, M. A. Khan, M. F. Mushtaq, R. A. Naqvi et al., “Intelligent breast cancer prediction
empowered with fusion and deep learning,” Computers, Materials and Continua, vol. 67, no. 1, pp. 1033–1049,
2021.
[36]   R. A. Naqvi, M. F. Mushtaq, N. A. Mian, M. A. Khan, M. A. Yousaf et al., “Coronavirus: A mild virus turned
CMC, 2022, vol.xx, no.xx xxxx

deadly infection,” Computers, Materials and Continua, vol. 67, no. 2, pp. 2631–2646, 2021.
[37]   F. Alhaidari, S. H. Almotiri, M. A. A. Ghamdi, M. A. Khan, A. Rehman et al., “Intelligent software-defined
network for cognitive routing optimization using deep extreme learning machine approach,” Computers,
Materials and Continua, vol. 67, no. 1, pp. 1269–1285, 2021.
[38]   M. W. Nadeem, M. A. A. Ghamdi, M. Hussain, M. A. Khan, K. M. Khan et al., “Brain tumor analysis empowered
with deep learning: A review, taxonomy, and future challenges,” Brain Sciences, vol. 10, no. 2, pp. 118–139,
2020.
[39]   Y. Masoudi-Sobhanzadeh, H. Motieghader and A. Masoudi-Nejad, “FeatureSelect: a software for feature
selection based on machine learning approaches,” BMC Bioinformatics, vol.20, no. 1, pp. 170, 2019.
[40]   M. Rahmaninia and P. Moradi, “OSFSMI: online stream feature selection method based on mutual
information,” Applied Soft Computing, vol. 68, pp. 733-746, 2018.
[41]   S. Pourbahrami, “Improving PSO Global Method for Feature Selection According to Iterations Global Search
and Chaotic Theory,” arXiv preprint arXiv:1811.08701.
[42]   L. Yu and H. Liu, “Feature selection for high-dimensional data: a fast correlation-based filter solution,” in
Proceedings of the 20th International Conference on Machine Learning (ICML-03), Washington DC, USA ,2003,
pp. 856-863.
[43]   A.L. Blum and P. Langley, “Relevance selection of relevant features and examples in machine learning,”
Artificial Intelligence, vol. 97, pp.245-271, 1997.
[44]   L.E. Raileanu and K. Stoffel, “Theoretical comparison between the GINI index and information gain criteria,”
Annals of Mathematics and Artificial Intelligence, vol. 41, pp. 77-93, 2004.
[45]   I. T. Jolliffe, “Principal Component Analysis”, 2nd Edition, Springer, New York, 2002
[46]   M.M. Kabir, M. Shahjahan and K. Murase, “A new local search based hybrid genetic algorithm for feature
selection”, Neurocomputing, vol. 74, no. 17, pp. 2914-2928, 2011.
[47]   L.-F. Chen, C.-T. Su, K.-H. Chen and P.-C Wang, “Particle swarm optimization for feature selection with
application in obstructive sleep apnea diagnosis,” Neural Computing and Applications, vol. 21, pp. 2087-2096,
2012.
[48]   B. Chen, L. Chen and Y. Chen, “Efficient ant colony optimization for image feature selection,” Signal Processing,
vol. 93, pp. 1566-1576, 2013.
[49]   C. Cortes and V. Vapnik, “Support-vector networks,” Machine Learning, vol. 20, no. 3, pp. 273–97, 1995.
[50]   K. Shaukat, S. Luo, S. Chen and D. Liu, “Cyber threat detection using machine learning techniques: a
performance evaluation perspective”. In International Conference on Cyber Warfare and Security
(ICCWS), Islamabad, Pakistan, 2020, pp. 1-6.  
[51]   C. J. Burges, “A tutorial on support vector machines for pattern recognition,” Data Mining and Knowledge
Discovery, vol. 2, no. 2, pp. 121-167, 1998.
[52]   J. Yim, “Introducing a decision tree-based indoor positioning technique,” Expert Systems with Applications,
vol. 34, no. 2, pp.1296-1302, 2008.
[53]   H. Sayoud, “Automatic Speaker Recognition-Connexionnist Approach,” PhD thesis, USTHB University, Algiers,
2003.
[54]   X. Solé, A. Ramisa and C. Torras, “Evaluation of random forests on large-scale classification problems using a
bag-of-visual-words representation,”. In Artificial Intelligence Research and Development. IOS Press, 2014
[55]   J. Han, J. Pei and M. Kamber, “Data Mining: Concepts and Techniques,” 3rd edition, Elsevier, 2011.
[56]   I. Cantürk and F. Karabiber, ‘‘A machine learning system for the diagnosis of Parkinson’s disease from speech
signals and its application to multiple speech signal types,’’ Arabian Journal for Science and Engineering, vol.
41, no. 12, pp. 5049–5059, 2016.
[57]   Y. Li, C. Zhang, Y. Jia, P. Wang, X. Zhang et al., ‘‘Simultaneous learning of speech feature and segment for
classification of Parkinson disease,’’ in Proceeding of IEEE 19th International Conference of e-Health
Networking, Application and Services (Healthcom), Dalian, China, Oct. 2017, pp. 1–6.
[58]   A. Benba, A. Jilbab and A. Hammouch, ‘‘Analysis of multiple types of voice recordings in cepstral domain using
MFCC for discriminating between patients with Parkinson’s disease and healthy people,’’ International Journal
of Speech Technology, vol. 19, no. 3, pp. 449–456, 2016.
[59]   A. Benba, A. Jilbab and A. Hammouch, ‘‘Using human factor cepstral coefficient on multiple types of voice
recordings for detecting patients with Parkinson’s disease,’’ IRBM, vol. 38, no. 6, pp. 346–351, 2017.