Available online at www.sciencedirect.

com ScienceDirect 
2212-0173 © 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license 
(http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of the organizing committee of ICETEST 
– 2015 doi: 10.1016/j.protcy.2016.05.128 
Procedia Technology 24 ( 2016 ) 1542 – 1549 

International Conference on Emerging Trends in Engineering, Science and Technology (ICETEST 

- 2015) 
A study on C.5 Decision Tree Classification Algorithm for Risk Predictions during Pregnancy Lakshmi.B.Na*, 
Dr.Indumathi.T.Sb, Dr.Nandini Ravic aPh.D Scholar, PG Research Centre, VIAT, VTU RRC, Muddenahalli, 
Chikkaballapura-562101, Karantaka, India bProfessor and Co-ordinator, PG Research Centre, VIAT, VTU RRC, Muddenahalli, 
Chikkaballapura-562101, Karnataka, India cDr.Nandini Ravi, MBBS, MD(Obs&Gyn), Dhruva Nursing Home, Hosakote, 
Bangalore, Karnataka, India 
Abstract 
Complication during pregnancy has turned out o be a major problem for women of today’s era. Pregnant women must be 
protected from these complications arising in period of gestation, a stage wherein every woman undergoes many 
physiological changes, sometimes inducing severe health problems leading to death of both mother and fetus. 
Technological interventions in the field of medical diagnosis can largely help to find a solution for this problem to 
protecting pregnant women, thus in turn reducing maternal and fetal mortalities to great extents. Decision Tree 
Classification method is a popularly used method whose algorithms are best suitable in medical diagnosis. C4.5 Decision 
Tree algorithm is one of the popular and effectively used classifier for pregnancy data classification in present study. The 
main aim of this paper is to pinpoint the importance of standardization of parameters selected for data collection in study, 
compare the results obtained from C4.5 classifier on both un-standardized and standardized datasets and analyse the 
performance of the C4.5 algorithm in terms of its prediction accuracy when applied on the created database from 
collected and standardized pregnancy data. © © 2016 2016 The The Authors. Authors.Published Published by by Elsevier 
Elsevier Ltd. Ltd. 
This is an open access article under the CC BY-NC-ND license 
(http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of the organizing committee of ICETEST 
– 2015. Peer-review under responsibility of the organizing committee of ICETEST – 2015 Keywords: pregnancy; learning 
models; C4.5 classification algorithm; pregnancy complications; abnormalities; standardized dataset; un- standardized dataset. 
1. Introduction 
Pregnancy  complications  leading  to  increased  rates  of  maternal  and  fetal  mortality  have  opened  a  new  horizon  for 
researchers to work on and provide a feasible solution to protect pregnant women from possible deaths. Pregnancy a delicate state 
in  each  woman’s life induces a variety changes in health leading to acute complications sometimes resulting in maternal and fetal 
mortality. The changes induced by physiological parameters are main reasons for an instance of complication to arise. 
* 
Lakshmi.B.N Tel: +91-8147949107 Email-id: [email protected] 
 
1543 B.N. Lakshmi et al. / Procedia Technology 24 ( 2016 ) 1542 – 1549 
Changes  in  physiological  parameters  if  un-noticed  by  medical  practitioners  or  pregnant  women  can  aggravate  to 
levels  of  emergency  situations.  When  an  emergency  situation  is  created  the  health  of  pregnant  woman  is 
complicated  and  difficult  to  normalize  and  in  such  instances  either  life  of  mother  or  child  is  lost  or  both  lives  are 
lost.  According  to  a  study,  each  day  around  800  women  die  from  different  causes  of  pregnancy  and  childbirth. 
Approximately  99%  of  all  these  maternal  deaths  occur  in  developing  countries.  A  study  in  2013  states  that  2,  89, 
000  pregnant  women  died  during  and  following  pregnancy  and  childbirth.  Almost  all  of  these  deaths  occurred  in 
low-resource  settings  and  most  of these could have been prevented [1][2]. Many of these causes can be prevented or 
controlled  if  these  pregnant  women  can  be  warned  before  the  situation  is  aggravated  to  complications  and  thus 
protect  the  pregnant  women  and  their  babies  from  any  further  complications  that  may  lead  to death. Some of these 
physiological  parameters  like  blood  pressure,  blood  glucose  level  and  weight  whose changes during pregnancy can 
lead  to  complications  can  be  identified  and  further  complications  induced  from  these  changes  can be prevented[1]. 
The aim of the paper is to use different types of databases for classification, to analyse and predict the level of risk in 
each  instance.  Here  in  this  paper  we  have  analysed  a  un-  standardized  database with all selected parameters for the 
study  and  also  analysed  a  database with parameters selected after standardization, a procedure that provides precise, 
valid,  reliable  and  accurate  parameters  for  study.  C4.5  Algorithm  is one of the most efficient, powerful and popular 
algorithm  of Decision Tree Induction method used for the task of classification of medical data, hence this algorithm 
is  selected  among  all the other algorithms of Decision Tree Induction method. This algorithm provided better results 
when  compared  with  other  algorithms  and  on  reviewing  many  works  from  researchers  it  was  found  that  C4.5 
outperforms  many  other  classification  algorithms  when  applied  on  medical  data.  The  second  section  briefs  about 
pregnancy  and  the  complications  induced  during  pregnancy.  The  third  section  presents  some  related  studies.  The 
fourth  section  provides  detailed  explanation  of  analysis  algorithm  C4.5  decision  tree  algorithm  and  its  working 
procedure,  used  to  predict  health  status  of  women  during  pregnancy.  The  last  section  presents  the  results  obtained 
from  the  analysis  on  both  the  standardized  data  and  un-standardized  data,  compares  the  results  obtained  from  the 
analysis  and  concludes  pinpointing  how  standardization  of  parameters  influence  the  results  obtained  from  C4.5 
algorithm for predicting health status from the collected pregnancy data. 
2. Literature Survey 
Jay  Gholap  aim  at  predicting  soil  fertility  class  using  decision  tree  algorithms  in  data  mining  and  focuses  on 
performance  tuning  of  J48(C4.5)  decision  tree  algorithm  with  the  help  of  meta  technologies  such  as  attribute 
selection  and  boosting.  [3].  Aniket  Ray  et  al  present  an  ensemble  classifier-hierarchical  committee of random tress 
that  uses  risk  factors  recorded  at  birth  in  order  to  predict  the  risk  of  developing  Retinopathy  of  prematurity.  [4]. 
Karpagavalli  S  et  al  present  the  implementation  of  3  supervised  learning  algorithms  C4.5  decision  tree  classifier, 
Naive  Bayes  and  Multilayer  perceptron  in  WEKA  environment  on  the  pre-operative  assessment  dataset.  The 
prediction  accuracy  of  classifiers  were  evaluated  using  10  fold  cross  validation  and  the  results  were  compared. 
[5].Mr.  Brijain  R  Patel  et  al  provide  focus  on  various  decision  tree  algorithms,  their  characteristic,  challenges, 
advantages  and  disadvantages.  [6].  T.Miranda Lakshmi et al compare and analyse the performance of ID3, C4.5 and 
CART  decision  tree  algorithms  using  qualitative  student  data.  [7].  Reza  Entezari-Maleki  et  al  have  compared  the 
efficacy  of  seven  classification  methods  Decision  Tree  (DT),  k-Nearest  Neighbor  (k-  NN),Logistic  Regression 
(LogR),  Naïve  Bayes  (NB),  C4.5,  Support  Vector  Machine  (SVM)  and  Linear  Classifier  (LC)  with  regard  to  the 
AUC  metric.  [8].  Sharareh  R.  Niakan  et  al  have  applied  and  compared  machine  learning  techniques  initially  to 
predict  the  outcome of tuberculosis therapy. After feature analysis six algorithms decision tree (DT), artificial neural 
network  (ANN),  logistic  regression  (LR),  radial  basis  function  (RBF),  Bayesian networks (BN), and support vector 
machine  (SVM)  were  developed  and  validated.  [9].  Masud  Karim  have  investigated  two  data  mining  techniques 
naïve  bayes  and  C4.5  decision  tree  algorithms and are implemented to predict whether a client will subscribe a term 
deposit. [10]. 
3. C4.5 Decision Tree Classification Algorithms 
 Classification  a  technique  most  suitable  in  the  area  of  medical  diagnosis  for  automatic  discovery  of  valid, 
novel,  unknown,  useful  and  understandable  information  from  very  big  databases  consisting  of  data  obtained  from 
several  applications.  Among  all  classification  methods  the  best  suitable classifier for the present study consisting of 
medical  data  is  C4.5  decision tree classification algorithm. The decision tree algorithms are powerful, popular, logic 
based, easily interpretable, straight forward and widely applicable for several problems 
 
1544 B.N. Lakshmi et al. / Procedia Technology 24 ( 2016 ) 1542 – 1549 
in  data  mining.  These  algorithms  are  off-the-shelf  algorithms  providing  very  good  performance  and  are  easily 
understandable. C4.5 Decision Tree Classification Algorithm was developed by Ross Quinlan as extension from ID3 
algorithm  also  developed  by  him.  These  classifiers  construct  a  decision  tree  as  a  learning  model  from  the  data 
samples.  The  divide  and  conquer  approach  is  followed  for  construction  of  decision  tree  models  using  a  measure 
called  information  gain  to  select  the  attribute  from  the  dataset  for  the  tree. Consider a possible test is to be selected 
with n outcomes which divides the data set L with training samples into subsets {L 
1 
, L 
2 
, L 
3 
.........L 
n 
}. Distribution of classes in L and its subsets L 
i 
is the only information available for tree construction. 
Considering P to be any set of samples, then freq (C 
i 
,P) are the total number of samples in P belonging to C 
i 
and |P| is 
denoted by the number of samples in P. The entropy for the set P is given by 
Info (P) = 
(1) 
Information  content  of  L  can  be  measured  by  computing  Info  (L),  the  total  information  content  of  L  can  be 
computed  once  L  is  divided  with  respect  to  the  outcomes  of  a  selected  attribute  say  z.  The  weighted  sum  of  the 
entropies of each subets gives the total information content of L. 
(2) 
The gain is given by 
Gain (z) = Info (L) 
(L) (3) 
Gives  information  gained  by  dividing  L  with  respect  to  the  test  on  z.  This  is done for the selection of attribute z 
with  highest  information  gain.  Three  base  cases  for  C4.5  algorithm  are  considered  (1)  If  all  samples  in  dataset 
belong  to  same  class  a  leaf  node  is  created  for  decision  tree  choosing  that  class.  (2)  If  no  information  gain  is 
provided  by any feature/attribute, a decision node is created high up the tree with expected value of the class. (3) If a 
class of an unseen instance encountered a decision node is created high up the tree with expected value of the class. 
4. Methodology 
The  section  of  methodology  provides  details  of  the  techniques  and  tools  used  for  data  collection,  selection  of 
respondents  and  methods  used  for  analysing  the  collected  data  instances.  The  section  briefs  about  the  scientific 
procedure followed for this present work to draw rational, logical and meaningful conclusions. 
4.1. Parameter Selection 
A set of attributes for the study are identified from the literature survey and suggestions of experts. Extensive 
literature survey is done to identify the problems faced by pregnant women in terms of changes induced by pregnant 
women and a set of parameters are selected. These parameters were subjected to further analysis by four doctoral 
committee members to suggest a set of valid parameters. Based on the suggestions from the committee members a 
total of twelve parameters are considered as shown in table 1. Table 1: Pregnant Women attribute and descriptions 
Sl. No 
Attribute Description Data type 
1  Age  Age  of  the  pregnant  women  in  years  Numeric  2  Pregnancy  Parity  Present  pregnancy  number  Number  3  History  of 
Pre-eclampsia Did the pregnant woman experience pre-eclampsia state in previous 
pregnancies (Yes/No) 
Textual 
4 History of Gestational 
Diabetes 
Did the pregnant woman experience gestational diabetes state in previous pregnancies (Yes/No) 
Textual 
5 Mother/Sister with pre- 
eclampsia 
Did the pregnant woman’s mother/sister experience pre-eclampsia state any of their pregnancies (Yes/No) 
Textual 
6 Mother/Sister with Did the pregnant woman’s mother/sister experience gestational Textual 
(L)= 
 
1545 B.N. Lakshmi et al. / Procedia Technology 24 ( 2016 ) 1542 – 1549 
Gestational Diabetes diabetes state any of their pregnancies (Yes/No) 7 State- The present health status of the pregnant 
woman 
(Hypertensive/Normal/Overweight/Underweight) 
Textual 
8  Trimester  Present  trimester of pregnancy (2/3) Numeric 9 Present Month Present month of pregnancy (4/5/6/7/8/9) Numeric 10 
Blood Pressure Blood Pressure recorded presently in mm/Hg Numeric 11 Presence of Gestational 
Diabetes 
Blood Sugar Levels recorded presently in mg/dl Numeric 
12 Weight Gain Present weight gain from the previous month in Kgs Numeric 
These  attributes  are  the  input  variables  identified  for  the  study  and  the  outcome  is  the  risk  level  in  pregnancy 
represented by tone of the status as Mild Risk, High Risk or No Risk. All the parameters are interlinked, abnormality 
in  one  parameter  will  lead  to  many  severe  complications  during  pregnancy.  Hence  efficiency  of  each  parameter  is 
important for health of pregnant women. 
4.2. Data Collection: 
Data  collection  is  carried  out  in  Bangalore  district,  Karnataka  state  for  present  study  and three medical centres 
from  Bangalore  district  are  identified  for  data  collection.  The  required  permissions  are  obtained  and  consent  of 
pregnant  women  visiting  for  monthly  consultations  are  also  obtained  for collecting data for the study. An interview 
schedule  was  developed  for  collection  of  that  incorporating  all  the  selected  attributes  of  the  study.  This  developed 
interview  schedule  was  used  for  data collection from pregnant women who visited the identified medical centres for 
monthly  check  up.  The  population  for  the  study  was  constituted  by  600  pregnant  women  from  who  visited  the 
medical  centres  for  check  up.  Figure  1  shows  the  interview  schedule  followed  for  data  collection.  Different 
measuring  devices  are  used  for  measuring  the  physiological  parameters  of  the  study.  The  values  collected  by  a 
medical expert are entered into the dataset. 
Fig 1: Interview schedule for data collection 
4.3. Pregnancy Dataset Creation: 
A pregnancy dataset is constituted by pregnancy data obtained from primary data collection using the interview 
schedule. 
• Un-standardized Pregnancy Dataset: The un-standardized pregnancy dataset is the pregnancy dataset obtained from 
interview  schedule.  The  dataset  is  checked  for  abnormal,  error-prone,  missing  and  irrelevant  data  samples.  If 
irrelevant  data  samples  are  identified  they  are  replaced  by  valid  values  or removed from the data set. The dataset is 
then  partitioned into training and testing datasets. The partitioning of dataset is done to subject the pregnancy dataset 
to analysis using C4.5 classification 
 
1546 B.N. Lakshmi et al. / Procedia Technology 24 ( 2016 ) 1542 – 1549 
algorithm. The dataset is randomly divided and around 370 data samples constitute the training data and remaining 
230 data samples constitute the test data. 
•  Standardized Pregnancy Dataset: Standardization is a statistical procedure followed by many researchers in various 
fields.  Standardization  of  study  parameters  provides  with  valid,  precise,  reliable,  meaningful  and  accurate  set  of 
parameters  leading  to  elevated  performance  by  the  selected  classifier.  The  procedure of standardization is followed 
since  pregnancy  is  a  delicate  medical  issue  and  parameters  influencing  pregnancy  can  be  better  determined  by 
medical  experts,  who  can  provide  a  reliable  set  of  parameters  used  for  the  study.  A  total  number of 40 judges who 
are  knowledgeable  and  experts  in  the  field  of  gynaecology  and  obstetrics  were  selected  in  random.  These  judges 
were  requested  to  give  their  opinion  regarding  suitability of the selected broad parameters for standardization. They 
were  requested  to  use  3  point  rating  relevancy  scale  as  suggested  by  Gulford(1954)  which  consisted  of  Most 
Relevant,  Relevant  and  Not  Relevant  scoring  used  for  responses  and  scored  2,1  and  0  respectively.  The  responses 
received from experts were subjected to analysis of Gulford(1954) 
relevancy component of ith component (R i 
) by using the formula suggested by 
(4) 
The parameters were subjected to relevancy test and the irrelevant parameters were removed from the list of 
parameters considered for the study. From a set of twelve parameters selected for the study six parameters were 
considered to be irrelevant by 40 medical experts and hence removed. A set of six parameters were considered for 
the study which had a relevancy percentage of more than 70. The remaining parameters of study are as listed along 
with their ranks in table 2. Table 2: Ranking order of parameters 
Parameter/Attribute Rank State 1 Blood Pressure 2 Blood Glucose Level 3 Weight 4 Trimester 5 Month 6 
Here  standardization  is  the  most  appropriate  method  for  parameter  selection  since  problem  considered  for 
analysis  is  a  sensitive and delicate medical issue and requires expertise in deciding the parameters for the study. The 
dataset  is  randomly  divided  and  around  370  data  samples  constitute  the  training  data  and  remaining  230  data 
samples constitute the test data. 
4.4. Data Analysis: 
Training  and  testing  datasets  are  analysed  using  C4.5  decision  tree  classification algorithm, the training dataset 
constituted  by  370  data  samples  is  used  to  construct  the  decision  tree  learning  model  and  testing  dataset  is used to 
evaluate the learning model constructed. 
•  C4.5  Classifier on Un-Standardized Dataset: Un-standardized pregnancy dataset is divided into training and testing 
datasets.  Training  dataset  has  370  data  samples  with  all  12  parametric  values  and  result,  the  risk-levels 
corresponding  to  each  data  sample.  Test  dataset  has 230 data samples with all 12 parametric values for each sample 
and  no  result  or  risk  level  corresponding  to  each  data  sample.  C4.5  classification  algorithm  is  applied  on  training 
dataset  to  build  the  decision  tree  learning  model  for  un-standardized  dataset.  On  obtaining  the  decision  tree  model 
test  dataset  is  used  to  analyse  the  performance  of  the  learning  model  in  terms  of  accuracy  in  prediction  of  risks 
during  pregnancy  from  un-standardized  pregnancy  dataset.  The  figure  3  shows  the application of C4.5 classifier on 
un-standardized training dataset and evaluation of constructed learning model using un-standardized test dataset. 
 
Fig 4:C4.5 on Standardized Dataset 
The purpose of this evaluation process is to determine the accuracy of the learning models in prediction and 
classification when subjected to two different types of databases. The knowledge obtained from this evaluation 
process helps in identifying the impact of parameters on the accuracy of an algorithm for predicting the risk induced 
during pregnancy in women. 5. Results 
The  learning  models  generated  by  C4.5  algorithm  are  easy  to  understand  and  require  no  domain  experts.  This 
technique  is  efficient,  powerful  and  popular  in  classifying  the  data  and  predicting  the  respective  risks  induced  in 
pregnancy.  The  study  also  highlights  a  slight  improvement  obtained  in  accuracy  when  standardized parameters are 
selected for analyzing through C4.5 classifier. 
Fig 3: C4.5 Classifier on Un-Standardized Dataset 
•  C4.5  classifier  on  Standardized  Dataset:  After  standardizing  parameters  of  study,  data  samples  corresponding  to 
those  parameters  constitute  the  standardized pregnancy dataset. Standardized pregnancy dataset is randomly divided 
into  370  training  and  230  testing  datasets.  Training  dataset  has  370  data  samples  with  selected 6 parametric values 
and  result,  the  risk-levels  corresponding  to  each  data  sample.  Test  dataset  has  230  data  samples  with  selected  6 
parametric  values  for  each  sample  and  no  result  or  risk level corresponding to each data sample. C4.5 classification 
algorithm  is  applied  on  training  dataset  to  build  the  decision  tree  learning  model  for  standardized  dataset.  On 
obtaining  the  decision  tree  model  test  dataset  is  used  to  analyse  the  performance  of  the  learning  model  in  terms of 
accuracy  in  prediction  of  risks  during  pregnancy  using  standardized  dataset.  The  figure  4  shows  the  application  of 
C4.5  classifier  on  standardized  training  dataset and evaluation of constructed learning model using standardized test 
dataset. 
1547 B.N. Lakshmi et al. / Procedia Technology 24 ( 2016 ) 1542 – 1549 
 
1548 B.N. Lakshmi et al. / Procedia Technology 24 ( 2016 ) 1542 – 1549 
Fig 5: Results of C4.5 on unstandardized test data 
Figure  5  shows  the  summary  obtained  from  the  analysis  of  C4.5  on  test  dataset.  From  230  data  samples  152 
data  samples  are  correctly  classified  or  the  corresponding  risk  levels  for data samples are predicted correcly and 78 
data samples are incorrectly classified or the risk levels are incorrectly predicted. 
Fig 6: Results of C4.5 on Standardized test data 
Figure  5  shows  the  summary  obtained  from  the  analysis  of  C4.5  on  test  dataset.  From  230  data  samples 152 
data  samples  are  correctly  classified  or  the  corresponding  risk  levels  for data samples are predicted correcly and 78 
data  samples  are incorrectly classified or the risk levels are incorrectly predicted. The extracted summary of analysis 
for  both  datasets  through learning model generated by C4.5 classifier using WEKA toolkit is as shown in the figures 
3  and  4.  The  accuracy  percentage  for  un-standardized  dataset  and  standardized  dataset  accounts  to  66.087%  and 
71.3043% respectively and error percentages are 33.913% and 28.6957% respectively. 
6. Conclusion 
It  is  evident  from  the  results  of  analysis  on  both  datasets  it  can  inferred  that  C4.5  decision  tree  classifier  has 
greater  potential  in  accuracy  for  predicting  the  risk  levels during pregnancy when applied on standardized data than 
on  un-standardized  data.  Figure  7  shows  the  graph  for  accuracy  and  error  obtained  by  C4.5  classifier  on  both  un- 
standardized  and  standardized  datasets.  The  graph  presents  that  C4.5  classifier  provides  better  performance  with 
standardization than with un-standardized dataset. 
 
Fig 7: Graph showing the performance of C4.5 classifier on pregnancy data 
In  this  study,  the  performance  of  C4.5  classifier  is  evaluated  for  accuracy  and  impact  of  standardization  on 
accuracy  is  understood  and  identified  for  providing  better  prediction  for  risk  in  pregnancy.  The  paper  provides 
significance  of  classification  and  prediction  in  mining  pregnancy  related  data.  Other  classification  techniques  can 
also  be  used  for  analysing pregnancy data but for this study C4.5 classifier is considered for this problem because of 
it  powerfulness,  popularity  and  efficiency  and  the  delicacy  of  the  pregnancy  problem.  Fig 7 gives the accuracy and 
error  percentages  by  C4.5  algorithm  when  applied  on  standardized  and  un-standardized  pregnancy  data.  Thus  as  a 
concluding  statement  C4.5  classifier  offers  better  performance if standardization is followed before analysis as seen 
from  results  obtained  from  both  the  datasets.  Hence,  from  this  study  the  paper  pinpoints  the  advantages offered by 
C4.5  decision  tree  classification  algorithm  for  prediction  based  health  monitoring  in  pregnant  women  and 
importance  of  standardization.  Research  is  further  carried  out  to  improve  the  accuracy  even  better  after  the 
standardization  procedure  to  provide  pregnant  women  a  precise  level  of  risk  to  give  them  a  safe  and  healthy 
pregnancy period. 
References 
[1]. United Nations Maternal Mortality Estimation Inter-agency Group, consisting of representatives from the World Health 
Organization (WHO), United Nations Children's Fund (UNICEF), the United Nations Population Fund (UNFPA), United Nations 
Population Division, the World Bank and Wikipedia. [2]. World Health Organization (November 2014). "Preterm birth Fact 
sheet N°363".who.int.Retrieved 6 Mar 2015. [3]. Applications of Machine Learning in Cancer Prediction and Prognosis by 
Joseph A. Cruz, David S. Wishart, Departments of Biological Science and Computing Science, University of Alberta Edmonton, 
AB, Canada T6G 2E8 in Cancer Informatics 2006: 2 59– 78 [4]. Decision Trees chapter 9 by Lior Rokach, Department of 
Industrial Engineerin, Tel-Aviv University and Oded Maimon, Department 
of Industrial Engineering, Tel-Aviv University [5]. Machine Learning Approach for Preoperative Anaesthetic Risk 
Prediction by Karpagavalli S1, Jamuna KS2, and Vijaya MS21,2 GR Govindarajulu School of Applied Computer Technology, 
PSGR Krishnammal College for Women, Coimbatore, India, Email: [email protected] in International Journal of Recent 
Trends in Engineering, Vol. 1, No. 2, May 2009 [6]. Predictive Data Mining for Medical Diagnosis: An Overview of Heart 
Disease Prediction by Jyoti Soni Ujma Ansari Dipesh Sharma, Student, M.Tech (CSE). Professor Reader, Raipur Institute of 
Technology, Raipur, Chhattisgarh, India in International Journal of Computer Applications (0975 – 8887) Volume 17– No.8, 
March 2011 [7]. Research of Knowledge Based Expert System Used in Maternity Diagnosis, Lu Binjie, Ph.D. of Donghua 
University, IT Department ˈ Renji Hospital affiliated Shanghai Jiaotong University School of Medicine, Shanghai, China, 
[email protected] in 2010 International Conference on Computer Application and System Modeling (ICCASM 2010, 
978-1-4244-7237-6/$26.00 C 2010 IEEE [8]. Supervised Machine Learning: A Review of Classification Techniques, S. B. 
Kotsiantis, Department of Computer Science and 
Technology, University of Peloponnese, Greece, End of Karaiskaki, 22100 , Tripolis GR. Informatica 31 (2007) 249-268 [9]. 
Performance Tuning of J48 Algorithm for Prediction of Soil Fertility, Jay Gholap, Dept. of Computer Engineering, College of 
Engineering, Pune, Maharashtra, India [10]. An Experimental Comparison of Naive Bayesian and Keyword-Based 
Anti-Spam Filtering with Personal E-mail Messages, Ion Androutsopoulos, John Koutsias, Konstantinos V. Chandrinos and 
Constantine D. Spyropoulos, Software and Knowledge Engineering Laboratory, Institute of Informatics and 
Telecommunications, National Centre for Scientific Research “Demokritos”, 153 10 Ag. Paraskevi, Athens, Greece 
1549 B.N. Lakshmi et al. / Procedia Technology 24 ( 2016 ) 1542 – 1549