Design and Analysis of Data Mining Based Prediction Model for Parkinson's disease

Abstract

Purpose: The purpose of this research paper is to develop a prediction model for Parkinson’s disease. There are many symptoms that lead to Parkinson’s disease such age- environmentalfactor, trembling in the legs, arms, hands, impaired speech articulation and production difficulties. In this research paper speech articulation of Parkinson’s disease affected people is considered for model formation and analyzes the model based on the
symptom of disease.

Methods: In proposed prediction model tree based classification model decision tree, ID3 and decision stumps
are used for training and testing the effectiveness of proposed prediction model. Here we also applied K –fold
cross validation technique for true prediction so that each record is sued for training and testing.

Results : In proposed model decision tree based our prediction model provide accuracy 85.08%, classification
error 14.92%, ID3 provide accuracy 75.33% ,classification error 24.67% and decision stumps based model
proved accuracy 83.55% and classification error 16.45%.

Conclusion:Proposed model based on Decision tree provide best result in comparison to other in terms of
parameters accuracy and classification error.

Full text

Design and Analysis of Data Mining Based
Prediction Model forParkinson’s disease
Chandrashekhar Azad1, Sanjay Jain2, Vijay Kumar Jha3
1Research Scholar, Department of CSE, Birla Institute of Technology, Mesra (Ranchi)
2Director, MICA Educational Company, Opposite Ranchi Club Gate, (Ranchi)
3Associate Professor, Department of CSE, Birla Institute of Technology, Mesra (Ranchi)
Abstract:
Purpose: The purpose of this research paper is to develop a prediction model for Parkinson’s disease. There are
many symptoms that lead to Parkinson’s disease such age- environmentalfactor, trembling in the legs, arms,
hands, impaired speech articulation and production difficulties. In this research paper speech articulation of
Parkinson’s disease affected people is considered for model formation and analyzes the model based on the
symptom of disease.
Methods: In proposed prediction model tree based classification model decision tree, ID3 and decision stumps
are used for training and testing the effectiveness of proposed prediction model. Here we also applied K –fold
cross validation technique for true prediction so that each record is sued for training and testing.
Results : In proposed model decision tree based our prediction model provide accuracy 85.08%, classification
error 14.92%, ID3 provide accuracy 75.33% ,classification error 24.67% and decision stumps based model
proved accuracy 83.55% and classification error 16.45%.
Conclusion:Proposed model based on Decision tree provide best result in comparison to other in terms of
parameters accuracy and classification error.
Keyword:Parkinsons, Data mining, Decision tree, ID3, Decision Stumps.
1. Introduction:
Neurons are the basic building blocks of thenervous system which incorporate the spinal brain and cord.
Neurons normally dont replaces or reproduce themselves. When neurons become damaged or die they cannot be
swapped by the body. Neurodegenerative diseases are Parkinson’s, Alzheimer’s, and Huntington’s disease [1,
2]. Today a lot of people are affected from neurodegenerative diseases like Parkinson disease, Alzheimer’s
disease, Arthritic disease, Prion disorders, Corticobasal degeneration, Progressive supranuclear palsy, Dementia
with Lewy bodies, Huntington’s disease , Motor neurone diseases , Huntington’s Disease, Spinal muscular
atrophy , Motor neurone diseases , etc. Neurodegenerative diseases are debilitating and incurable conditions
that result in progressive degeneration of nerve cells,and it causes problemsinmental functioning and
movement. In Scotland 120 to 230 people in per 100,000 are affected with Parkinson’s disease, While
population of the Scotland remains stable. In the next 25 years Parkinson’s disease affected people may
increase by 25-30 % [3]. Parkinson’s disease is the 2nd most neurodegenerative diseases. It is identified by
progressive loss in control of muscle, and it leads to trembling ofhead and limbs, and at restslowness,impaired
balance and stiffness. Day by day symptoms worsen and the affected people may have difficulty in
walking,talking and also may in complete simple tasks. In USapproximate 1 million people affected with
Parkinson's disease and around theworld approximate 5 million people affected with it. Most of the 60 years or
olderage people are affected with Parkinson’s disease, it is found approximately 1% in 60 age group and
approximate 4% in the age group of 80 years. Since overall life expectationrising and maythe number of people
with Parkinson's disease will increase in the near future. Adult-onset PD(Parkinson’s Disease) is common,
early-onset is in 21-40 years, and juvenile-onset PD before age 21. The Parkinson's disease is date back as far
as approximate 5000 BC and Indian civilization termed it as the Kampavata and the they use the seeds of a plant
that contain therapeutic levels for treatment. Parkinson's disease was discovered by James Parkinsonin 1817 as
"shaking palsy"[4]. Near about 1 million adults in USA are affected withPD and over 60,000 people diagnosed
every year. In USA, According to Parkinson's disease Foundation, $25 billion annually expend in thePD and
average annual medication costs is in the range of $2,500 to $10,000. In United Kingdom 1 in every 500
people is affected with PD and about 10 million people in world. A male have 50% higher risk than a female in
developing Parkinson's disease.In the many of the cases, the symptoms are appear after the age 50 and approx
4-5% of cases in younger than 40 years age group[4,5,6].
Chandrashekhar Azad et al. / International Journal of Computer Science Engineering (IJCSE)
ISSN : 2319-7323
Vol. 3 No.03 May 2014
181

Figure 1. Parkinson's Diseases Affected
Symptoms of Parkinson’s Disease?
 slowness of movement, slowed motion(Bradykinesia)
 Resting tremor
 muscle stiffness
 Posture and balance
 The arms may not swing when walking
 Swallowing difficulties
 Speech problems
 Loss of facial expression
Possible complications of Parkinson’s disease
 Depression
 Sleeping problem
 Urinary incontinence or retention
 Constipation
 Thinking difficulties
Risk Factors:
 Age
 Gender
 Family History
 Race and Ethnicity
2. Applications of data mining for classification:
Decision tree
Decision tree is supervised learning technique in data mining[23,24,25]and it is used for classification which
maps unlabeled records to a target class based on the learned model in other word we can termed it as
classification trees. In decision tree or classification tree leaf nodes represent class labels and branches represent
the test outcome of the features. In the decision analysis process it is used to visually represent the decisions.
The goal of decision tree classifier is to produce a predictionmodel with the historical records termed as a
training set and the learned model predicts the target value of a given input variables set. A decision tree can be
trained by splitting the dataset into subsets based on an attribute value test and the training process is repeated
on each subset in a recursive manner. The training iscompleted when splitting of attribute remain no longer to
addin the predictions.
Data format:,   ,
,
…….
,
The Y is the target class that we are trying do classify. The vector  ,  , , …  is set of input variables..
Chandrashekhar Azad et al. / International Journal of Computer Science Engineering (IJCSE)
ISSN : 2319-7323
Vol. 3 No.03 May 2014
182

Attribute Selection Measures:
Attribute selection for decision tree is heuristic for selecting the splitting criterion that best separatesthe given
data set into individual classes. Splitting of dataset into smaller partitions based on the outcomes of the splitting
criteria, each partitions are pure. The best splitting criteria is the one that have the best value set to split.
Attribute selection is alsotermed as splitting rules,sincethe splitting rules of criteria determine how the given
tuples to be split.The attribute selection criteria give the ranking to each attribute based the given training set.
The attributethathave the best score for the splitting,is chosen for splitting the given dataset. If splitting attribute
is continuousvalued or if restricted to binary trees, then, either a split point or a splitting subset must be
determined as a part of the splitting criteria. Tree node created for partition in training set is labeled with the
outcome of the slitting criteria and branches are extended for each condition of the splitting criteria, and the
given training tuples are partitioned accordingly. The attribute selection measures are: information gain, gain
ratio, and Gini index.
ID3:
In decision tree learning, Iterative Dichotomiser (ID3) algorithm used to generate a classification tree or
decision tree from the given training set. ID3 is the successor ofC4.5 algorithm, and is typically used in the
machine learning or data mining for classification.
Decision stumps:
A decision stump is machine learningclassification model, it consist of a one-level decision tree. Decision stump
is a decision tree with one internal node is called root node and is immediately connected to the leaf nodes or
terminal node. It makesthe classification based on the value of a single input. Decision stump is also called 1-
rules.For nominal attribute decision stump classifier build a stump which hold a leaf for each possible attribute
value or a stump having two leaves, one corresponds to some chosen category, and the other leaf to all the other
categories.
Attributes having continuous values, some threshold attribute value is choosen, and the decision stump contains
two terminals for values below and above the defined threshold. In this a missing value is treated as another
category.
Dataset:
The Parkinson’s disease data set is taken from UCI repository. This is built up ofdata of 31 people, 23 with
Parkinson's disease (PD) and rest are healthy. This data set contains the 197 instance and each attribute have the
real values. The target of the dataset is to distinguish Parkinson’s disease affected from those with non
parkinsons diseases affected, in the dataset 0 is labeled for healthy and 1 for Parkinson’s disease. The
Parkinson’s disease dataset was created by Max Little, University of Oxford [22].Statistics of the dataset is
given below:
 Dataset Characteristics : Multivariate
 Attribute Characteristics : Real
 Number of Instances : 197
 Number of attribuite : 23
 Missing Values : None
 Area : life
Chandrashekhar Azad et al. / International Journal of Computer Science Engineering (IJCSE)
ISSN : 2319-7323
Vol. 3 No.03 May 2014
183

Table1: Dataset description
Role Index Attribute Name Type Description
label 23 Status binominal Health Status P for parkinsons and H for Healthy
regular 0 Name polynominal ASCII subject name and recording number
regular 1 MDVP_Fo_Hz real Average vocal fundamental frequency
regular 2 MDVP_Fhi_Hz real Maximum vocal fundamental frequency
regular 3 MDVP_Flo_Hz real Minimum vocal fundamental frequency
regular 4 MDVP_Jitter real Kay Pentax MDVP jitter as percentage
regular 5 MDVP_Jitter_Abs real Kay Pentax MDVP absolute jitter in microseconds
regular 6 MDVP_RAP real Key Pentax MDVP Relative Amplitude
Perturbation
regular 7 MDVP_PPQ real Kay Pentax MDVP five-point Period Perturbation
Quotient
regular 8 Jitter_DDP real Average absolute difference of differences
between cycles, divided by the average period
regular 9 MDVP_Shimmer real Key Pentax MDVP local shimmer
regular 10 MDVP_Shimmer_dB real Key Pentax MDVP local shimmer in decibels
regular 11 Shimmer_APQ3 real 3 Point Amplitude Perturbation Quotient
regular 12 Shimmer_APQ5 real 5 Point Amplitude Perturbation Quotient
regular 13 MDVP_APQ real Kay Pentax MDVP eleven-point Amplitude
Perturbation Quotient
regular 14 Shimmer_DDA real Average absolute difference between consecutive
differences between the amplitude of consecutive
periods
regular 15 NHR real Noise to Harmonic Ratio
regular 16 HNR real Harmonics to Noise Ratio
regular 17 RPDE real Recurrence Period Density Entropy
regular 18 DFA real Detrended Fluctuation Analysis
regular 19 spread1 real Non Linear measure of fundamental frequency
regular 20 spread2 real Non Linear measure of fundamental frequency
regular 21 D2 real Correlation Dimension
regular 22 PPE real Pitch Period Entropy
Chandrashekhar Azad et al. / International Journal of Computer Science Engineering (IJCSE)
ISSN : 2319-7323
Vol. 3 No.03 May 2014
184

3. Proposed Prediction Model:
Figure 2. Proposed Prediction Model
Steps in Prediction Process:
Step 0: Start
Step 1: Load data set
Step 2: Model Creation using Training set
Step 3: Testing Model using validation set
Step 4: Performance analysis
Step 5: Selection Of best Model
Step 6: Stop
4. Experiments and Discussion:
This section describe our experiment result, Experiment is carried out using rapid miner on a system having
intel i5 3rd generation processor, 4 Gb RAM, 500 GB harddisk, Windows 7 ultimate operating sytem . In the
first study, three diffent types of classification algorithms are used to predict a person is either helthy or
parkinsons affected. In this experiment used classification algorithms are decision tree, ID3 and decision stump
and for parameter we used accuracy and classification error. In the process of classification or prediction of
helathey or parkinsons we taken dataset from UCI repository and using the three well known classification
method decision tree, ID3 and decision stump to trained the model . Here we used 10 – Fold cross validation
technique is used to complete our experiment, reason behind choosing 10 fold cross validation for traing &
testing the effectiveness of proposed prediction model is for unbiased predication. In k – fold(Here 10-fold)
cross validation technique entire dataset is divided into k parts and K-1 parts are used for training and kth part is
taken as testing set , this proess is repeating k times so that each part is taken as testing set.
Figure 3. K fold cross validation
Chandrashekhar Azad et al. / International Journal of Computer Science Engineering (IJCSE)
ISSN : 2319-7323
Vol. 3 No.03 May 2014
185

Table 2. Classification Results
F Decision Tree ID3 Random Tree
CM Acc CE CM Acc CE CM Acc CE
1 12 0 85.00 15.00 15 5 75.00 25.00 15 3 85.00 15.00
3 5 0 0 0 2
2 11 1 78.95 21.05 14 5 73.68 26.32 14 5 73.68 26.32
3 4 0 0 0 0
3 15 3 85.00 15.00 15 5 75.00 25.00 15 4 80.00 20.00
0 2 0 0 0 1
4 13 2 84.21 15.79 14 5 73.68 26.32 14 3 84.21 15.79
1 3 0 0 0 2
5 13 1 85.00 15.00 15 5 75.00 25.00 15 2 90.00 10.00
2 4 0 0 0 3
6 14 2 89.47 10.53 14 5 73.68 26.32 14 3 84.21 15.79
0 3 0 0 0 2
7 15 2 90.00 10.00 15 5 75.00 25.00 15 3 85.00 15.00
0 3 0 0 0 2
8 13 4 73.68 26.32 14 5 73.68 26.32 14 4 78.95 21.05
1 1 0 0 0 1
9 15 2 89.47 10.53 15 4 78.95 21.05 15 2 89.47 10.53
0 2 0 0 0 2
10 15 1 90.00 10.00 16 4 80.00 20.00 15 2 85.00 15.00
1 3 0 0 1 2
Mean 85.08 14.92 75.37 24.63 83.55 16.45
Figure 4. Decision Tree
Chandrashekhar Azad et al. / International Journal of Computer Science Engineering (IJCSE)
ISSN : 2319-7323
Vol. 3 No.03 May 2014
186

Figure 5. ID3 Tree
Figure 6. Tree (Decision Stump )
Figure 7. Accuracy by Fold
Chandrashekhar Azad et al. / International Journal of Computer Science Engineering (IJCSE)
ISSN : 2319-7323
Vol. 3 No.03 May 2014
187

Figure 8. Classification Error by fold
Figure 9. Mean Accuracy
Figure 10. Mean Classification Error
In this paper table 1 provide the description of dataset used in this for training and testing the effectiveness of
the proposed model. This table 1 it includes name of attributes or features type of feature, description of the
attribute, table index or position of the attribute in the training and testing file. Table 2 shows the results of the
experiments carried out using the decision table, ID3 and decision stumps. In the table 2 ACC stands for
accuracy and CE stands for classification error. Figure 1 shows the posture a person who is Parkinson diseases
affected. In Figure 2 it describe proposed prediction model for parkinsons diseases. In Figure 3 we showed
pictorial representation of K fold cross validation. Figure 4, 5,6 describe the Tree generated through
classification models decision tree ,ID3 and decision stumps respectively. Figure 7 and figure 8 describe the
accuracy and classification error by 10 fold cross validation and the figure 9 and figure 10 shows the mean
accuracy and classification error . Easily we can say using the graph, decision tree provide the best result.
Chandrashekhar Azad et al. / International Journal of Computer Science Engineering (IJCSE)
ISSN : 2319-7323
Vol. 3 No.03 May 2014
188

5. Conclusion:
In this paper prediction of parkinsons disease paper we proposed a prediction model using data mining method,
for predictions we used decision tree , ID3, and decision stumps classification algorithms. Dataset of parkinsons
disease is used in this paper is taken from UCI repository. The dataset of Parkinson’s disease iscomposed of
biomedical voice measurements of 31 people, with 23 PD affected. Each column in thetable is a particular voice
measure, and each row corresponds one of 195 voice recording from these individuals. Parkinson’s
diseaseaffected data is labeled with P and healthy is with H. For experimentsK fold cross validation method
isused with different classifiers as well as classification accuracy and error. The meanresults show in terms of
classification accuracyerror, Decision tree has here performed very well i.e., accuracy 85.08 and classification
error 14.92 and model ID3 performed worst, it gave accuracy 75.33 and classification error 24.63. There are
different symptoms that lead to the Parkinson’s disease are age and environmental factor, trembling in the legs,
arms, hands, impaired speecharticulation and production difficulties. In this research paper speech articulation
of Parkinson’s disease affected people is considered for model formation and analyzes the modelbased on the
symptom of disease.
References:
[1] Alexis Elbaz,James H. Bower, Brett J. Peterson, Demetrius M. Maraganore, Shannon K. McDonnell, J. Eric Ahlskog, Daniel J.
Schaid, Walter A. Rocca, “Survival Study of Parkinson Disease in Olmsted County, Minnesota “, Arch Neurol. Vol. 60:91-96, 2003
[2] Parkinson, James (1817),” An essay on the shaking palsy”
[3] Newman EJ, Grosset KA, Grosset DG. Geographical difference in Parkinson’s disease prevalence within West Scotland. Mov Disord
2009;24(3):401-6
[4] http://www.medicinenet.com/parkinsons_disease/article.htm
[5] http://www.medicalnewstoday.com/info/parkinsons-disease/
[6] http://www.umm.edu/patiented/articles/who_gets_parkinsons_disease_000051_3.htm
[7] Rajesh Pahwa, Kelly E. Lyons, William C. Roller,"Handbook of Parkinson's Disease" ,Third Edition
[8] William Dauer,Serge Przedborski,"Parkinson’s Disease: Mechanisms and Models “,Neuron, Vol. 39, 889–909, September 11, 2003,
Copyright 2003 by Cell Press
[9] Sanjay Pandey,"Parkinson`s Disease : Recent Advances",JAPI • june 2012 • VOL. 60
[10] James Parkinson,"An Essay on the Shaking Palsy",J Neuropsychiatry Clin Neurosci 14:2, Spring 2002.
[11] Stanley Fahn and the Parkinson Study Group,"Does levodopa slow or hasten the rate of progression of Parkinson’s disease?",Neurol
(2005) 252 [Suppl 4]: IV/37–IV/42
[12] Oliver Riedel et al., "Cognitive impairment in 873 patients with idiopathic Parkinson’s disease Results from the German Study on
Epidemiology of Parkinson’s Disease with Dementia (GEPAD)",J Neurol (2008) 255:255–264
[13] Nathan Pankratz et al.,"Genomewide association study for susceptibility genes contributing to familial Parkinson disease ",Hum Genet
(2009) 124:593–605,Springer-Verlag 2008.
[14] Keiko Tanaka et al. , " Occupational risk factors for Parkinson’s disease:a case-control study in Japan",BMC Neurology 2011
[15] Calvin Yu-Chian Chen, "Mechanism of BAG1 repair on Parkinson’s diseaselinked DJ1 mutation ",Journal of Biomolecular Structure
and Dynamics , Taylor & fransis,Vol. 30, No. 1, 2012, 1–12
[16] Antonio Del Sol Mesa , "151 Network inference and analysis of Parkinson’s disease",Journal of Biomolecular Structure and Dynamics
Vol. 31, Supplement, 2013
[17] Audrey McKinlay,Randolph C. Grace "Characteristic of Cognitive Decline in Parkinson’s Disease: A 1-Year Follow-Up",APPLIED
NEUROPSYCHOLOGY, 18: 269–277, 2011
[18] C. W. Olanow and W. G. Tatton,"Etiology and pathogenesis of Parkinson’s disease",Annu. Rev. Neurosci. 1999. 22:123–44
[19] Marco Aurélio M. Freire1, and José Ronaldo Santos,"Parkinson’s disease: general features, effects of, levodopa treatment and future
directions",Frontiers in Neuroanatomy,November 2010,Volume 4.
[20] Yadav G, Kumar Y, Sahoo G. "Predicationof Parkinson’s disease using data mining methods: A comparative analysis of tree,
statistical, and support vector machine classifiers."Indian J Med Sci 2011;65:231-42
[21] Chrish zarow etal, “Neuronal Loss Is Greater in the Locus Coeruleus Than Nucleus Basalis and Substantia nigra in Alzheimer and
Parkinsons Diseases.”, American medical association, 2003
[22] Max A. Little, Patrick E. McSharry, Eric J. Hunter, Lorraine O. Ramig (2008), 'Suitability of dysphonia measurements for
telemonitoring of Parkinson's disease', IEEE Transactions on Biomedical Engineering.
Chandrashekhar Azad et al. / International Journal of Computer Science Engineering (IJCSE)
ISSN : 2319-7323
Vol. 3 No.03 May 2014
189