Comparative Analysis of Early Detection of Hypothyroidism using Machine Learning Techniques

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 09 Issue: 08 | Aug 2022 www.irjet.net p-ISSN: 2395-0072
© 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 230
Comparative Analysis of Early Detection of Hypothyroidism using
Machine Learning Techniques
Ranjitha B1, K R Sumana2
1PG Student, The National Institute of Engineering, Mysuru, Karnataka, India
2Assistant Professor, Mysuru, Karnataka, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - The diagnosis of health conditions and proper
treatment of disease at an early stage is one of the most
challenging tasks in the healthcare field. Hypothyroidism is a
type of thyroid disease. Thyroid glands are located in the
middle of our necks. It has a butterfly shape and is small in
size. People with hypothyroidism do not produce enough
thyroid hormone to keep their bodies functioning normally.
The thyroid gland may be involved in several conditionseither
directly or indirectly. Damage to the thyroid gland and
inflammation are the causes of hypothyroidism. Low thyroid
hormone levels cause the body’s functions to slow down,
leading to general symptomslikefatness, lowpulse, increasein
cold sensitiveness, neck swelling, dry skin, hands symptom,
hair drawback, serious emission periods. The purpose of this
project is to predict the Hypothyroidism disease at the early
stage. Nowadays, machine learning has become an incredibly
popular way to detect various diseases. Machine learning is
used to detect disease at an early stage with greater accuracy.
This Project uses KNN, Random Forest(RF) and XGB
algorithms to predict the hypothyroidism disease at the early
stage.
Key Words: Thyroid disease, Hypothyroidism, KNN,
Random Forest ,XGB
1.INTRODUCTION
Thyroid is one of our glands, whichmakehormones.Thyroid
hormones control the rate of numerous conditioning in our
body. It secretes a few chemicals that are blended in with
blood and excursion across the body to control modling.
There are two primary thyroid chemicals Triiodothyronine(
T3) and Thyroxin( T4). Thesetwochemicalsaresignificantly
answerable for keeping up with the energy in our bodies.
The two main types of thyroidconditionareHypothyroidism
and Hyperthyroidism. Hypothyroidism is caused when the
gland releases low situations ofthyroidhormone.Symptoms
of an underactive thyroid(hypothyroidism) can include
Feeling tired, Gaining weight, passing obliviousness, Having
frequent and heavy menstrual ages, Having dry and coarse
hair, Having a coarse voice. Foodsthataffectthyroidaretofu,
tempeh, edamame sap, soy milk, etc. The potables coffee,
green tea, and alcohol — these potables may irritate our
thyroid gland. People who worked 53– 83 hours per week
were shown to have a higher rate of hypothyroidism than
those who worked 36–42 hours per week. The night shift
work might be associated with the threat of subclinical
hypothyroidism, and that this threat increased with longer
employment as a night shift worker. While sleeping further
than eight hours per day may increase the threat of both
hyperactive and underactive thyroid function. Thyroid
disease affect an estimated 200 millionpeople worldwide.In
India there are 42 Million people have thyroid diseases and
Hypothyroidism is utmost of the common thyroidcomplaint
in India.
2. RELATED WORK
The authors [1] in this article applied the classification
(KNN) and prediction model (decision tree) to the thyroid
dataset to accurately predict new patient entry. The KNN
algorithm is used to classify thyroid disorders with related
prioritized symptoms. Artificial Neural Network, support
vector machine, Naive Bayes and KNearest Neighbor arethe
important modes applied tothepredictionofthyroiddisease
and the results show that the K-nearest neighboraccuracyis
better than any other thyroid disease detection technique.
[2]utilized information mining calculations, for example,
KNN, Naive Bayes, Support Vector Machine for the
concentrate in this paper. The after effects of these
arrangement techniques depend on the precision and
execution of the model. For the given dataset, SVM accuracy
is 0.82, Naive Bayes accuracy is 0.83 and KNN accuracy is
0.85. [3] Utilizes calculations like KNN, Random Forest,
Naive Bayes, and ANN. KNN with Random Forest exhibited
improved results with a precision of 94.8 percent when
contrasted with the complete outcomes with four classifiers
on the equivalent dataset. Utilizing decision tree algorithm,
random forest algorithm, supportvectormachinealgorithm,
logistic regression and multilayerfeedforwardalgorithm[4].
After doing a comparative analysis to identify the prediction
algorithm that produces the most precise and accurate
results, it can be said that the decisiontreealgorithmdoesso
with a 99.46 percent accuracy rate and precision 0.99. The
informational collections for the thyroid sicknesses have
been had from the UCI website. The Machine Learning
Algorithms like Artificial Neural Network, Support Vector
Machine, Decision Tree, K-Nearest Neighbor are utilized to
arrange and anticipate the exactness. Thyroid infection
prescient models which require least number of boundaries
of an individual to analyze thyroid illnessandsetsasideboth
cash and season of the patient. [5]This paper studies on
thyroid disease and apply some algorithms to test
performance study on mentioned algorithms. ANN-97.50,

KNN-98.62, SVM-99.63 and DT -75.76. Without a doubt,
many thyroid diseases have been successfully diagnosed by
experts worldwide. However, it is recommended that
patients employ fewer diagnostic criteria when seeking a
thyroid condition diagnosis. With more characteristics, a
patient must undertake testing, which is both time- and
money-consuming. In order to save patients' time and
money, it is vital to develop algorithms and predictive
models for thyroid disease that only require a few factors
provided by the patient to detect the issue.
3. PROPOSED SYSTEM
Currently, machine learning hascomeanimmenselypopular
medium for detecting various diseases. It's veritably
accessible and effective to presume conditions using
machine learning ways. We've used Machine learning
algorithms such as KNN, Random Forest( RF) and XGB.
Eventually derived that these algorithms helps us to attain
better delicacy.
4. METHODOLOGY
Machine learning, a subfield of man-made brainpower,
empowers PCs to "learn" for themselves from preparing
information and work on over the long run without being
expressly customized. Algorithms can make their own
forecasts by recognizing designs in the information and
gaining from them. Machine Learning Algorithms like
Random Forest, K-Nearest Neighbor and XGBoost are
utilized to anticipate the hypothyroidism in beginning
phases.
4.1 ALGORITHMS
[1] Random Forest
A supervised learning algorithm is random forest. However,
classification issues still account for a largeportionofitsuse.
As we all know, trees make up a forest, and more trees equal
a more healthy forest. Additionally, the random forest
algorithm builds decision trees from data samples, extracts
forecasts from each one, and then uses voting to select the
fashionable outcome. An ensemble system reduces over-
fitting by combining the results, making it superior to a
single decision tree.
[2] KNN
The KNN represents "K-Nearest Neighbor". The algorithm
can be utilized to break both section and regression issue
proclamations. The picture "K" stands for the number of
nearest neighbours to anotherambiguousvariablethatmust
be anticipated or organised. KNN operates by taking a
chance on the distances between a question and each
embodiment in the data, selecting the predetermined
number representations ( K) nearest to the inquiry,likewise
votes in favor of the most successive data.
[3] XGB
Extreme Gradient Boosting, or XGBoost, is an idea putout by
University of Washington specialists. Loads are critical in
XGBoost. Every free factor is given a load prior to being
taken care of into the choice tree that gauges results.Factors
that the tree inaccurately anticipated are given more weight
prior to being set into the subsequent choice tree. These
unmistakableclassifiers/indicatorsarethenjoinedtodeliver
a hearty and precise model.
4. DATASET
The dataset are collected from the kaggle website. In this
project we used some attributes to get patient details like
Age of the patient, Sex-Patient Male/Female andsomeofthe
clinical details like Thyroxin,
Antithyroid_medication,Goiter,Psych,T3, TT4, T4U, FTI.
TABLE -1: Dataset List
Attribute name Description
Age Age of the patient
Sex Male/Female
Thyroxin Clinical Test True/False
Antithyroid_medication Clinical Test True/False
Goiter Clinical Test True/False
Hypopituitary Clinical Test True/False
psych Clinical Test True/False
T3 Clinical Test Value
TT4 Clinical Test Value
T4U Clinical Test Value
FTI Clinical Test Value
Class
negative,
compansated_hypothyroid,
primary_hypothyroid,
secondary_hypothyroid

5. SYSTEM ARCHITECTURE
FIG -1: System architecture
Fig. No. 1 shows the system architecture, First collect the
dataset from the website and dataset is stored in the
database. Taken by the website are larger dataset that must
be sample to balance. Then Clean the missing values and
select the important attributes. Then apply feature selection
and train the model. Then divide into testing and training
model, apply algorithms i.e., knn, rf, xgb for getting better
accuracy and finally we get result as early prediction of
hypothyroidism.
6. EVALUATION AND TESTING
6.1 DATA PREPROCESSING
The data collected from the website are not formatted
clearly. So, we should check for missing values in the list.
Drop the unnecessary attributes and keep necessary
attributes. Then taking care about catagorical data i.e.,inour
dataset-class attribute had negative,
compensated_hypothyroid, primary_hypothyroid
secondary_hypothyroid to find hypothythyroidism in early
stages these classes are used. Then change object type to
numerical type data. To apply machine learning algorithms,
data must be split into training and testing.
6.2 TRAINING SET
To check training and testing, we should separate
independent and dependent variables as X and Y and check
for imbalanced data. By using Kmenas clustering will
clusters the data. The class attributes are negative,
compensated_hypothyroid primary_hypothyroid
secondary_hypothyroid and the values as 0, 1, 2, 3
respectively. Then split into traing and testing data. Then
create model as x_tarin,y_train, x_test and y_test.
Then, apply machine learning algorithms-A model called
random forest is built using several decision trees. When
constructing trees and breaking nodes, the model randomly
selects samples from the training data points.Applyxtarin,y
train, x test, and y test for the Random Forest Classifier to
obtain accuracy for this algorithm.
Then, XGB works as load prior to being taken care of intothe
choice tree that gauges results. Factors that the tree
inaccurately anticipated are given more weight prior to
being set into the subsequent choice tree. These
unmistakableclassifiers/indicatorsarethenjoinedtodeliver
a hearty and precise model and apply xgb model to
x_tarin,y_train, x_test and y_test to get accuracy for this
algorithm.
7. RESULT
After the implementation using KNN, Random forest and
XGBoost Machine learningalgorithms,andall oftheclassifier
results are compared. Then evaluated the results based on
the early stages like negative, compensated_hypothyroid,
primary_hypothyroid, secondary_hypothyroid. Fig.No. 2
Shows the result of the hypothyroidism dataset. Classes are
negative, compensated_hypothyroid, primary_hypothyroid,
secondary_hypothyroid and the values are 0, 1, 2, 3
respectively. Here x-axis defines the class, in which class the
patients are in highest count according to given dataset
(negative, compensated_hypothyroid,primary_hypothyroid,
secondary_hypothyroid ) and y-axis defines the count of the
patients(0, 500, ….3500) data. The dataset values arealmost
compensated_hypothyroid patients only indicatesthat early
stage of hypothyroidism patients are found as shown in the
chart-1 below and the accuracy for these algorithms are
shown in table below.
CHART -1: Number of Patient’s having disease at earlystage

TABLE -2 Comparison of Algorithms
8. CONCLUSIONS
In our discoveries, we have seen that KNN, Random Forest
and XGBoost Algorithms are utilized to assists us with
foreseeing hypothyroidism in the beginning phase by
utilizing a continuousdataset.Intheproposedframework its
seen that an impediment of information to work with. At
order to find a better solution and be better prepared to
predict illness in its crucial stage, we will need to work with
a larger dataset in the future. We also hope that morepeople
from our country will be interested in dealing with this
illness. Trust that will enable citizens of our nation to
maintain a healthy society.
REFERENCES
[1] Prediction Of Thyroid Disease Based On Classification
Using Hierarchical Structure By Charan R, Akash Yadav M,
Aprameya N Katti, Mohith P Global Academy Of Technology,
Bengaluru Karnataka 560098, India
[2] Thyroid Disease Prediction Using Feature Selection And
Machine Learning ClassifiersByDr.DayanandJamkhandikar
, Neethi Priya
[3] Empirical Method For Thyroid Disease Classification
Using A Machine Learning Approach By Tahir Alyas ,
Muhammad Hamid , 2 Khalid Alissa, Tauqeer Faiz, Nadia
Tabassum , And Aqeel Ahmad
[4] Thyroid Prediction Using Machine Learning Techniques
By Sagar Raisinghani, Rahul Shamdasani , Mahima Motwani,
Amit Bahreja , And Priya Raghavan Nair Lalitha Department
Of Computer Engineering, University Of Mumbai,
Vivekanand Education Society’s Institute Of Technology,
Mumbai, India
[5] Interactive Thyroid Disease Prediction System Using
Machine Learning Technique By Ankita Tyagi ,Ritika Mehra,
Computer Applications, Aditya Saxena, Computer Science
And Engineering Dit University Dehradun, India
Algorithms Accuracy Score
Time
Efficiency
Random-Forest
Classifier
88.5% 0.897 0.3ms
XGBoost
Classifier
87.8% 0.902 1.5ms

Comparative Analysis of Early Detection of Hypothyroidism using Machine Learning Techniques

More Related Content

Similar to Comparative Analysis of Early Detection of Hypothyroidism using Machine Learning Techniques (20)

More from IRJET Journal (20)

Recently uploaded (20)

Comparative Analysis of Early Detection of Hypothyroidism using Machine Learning Techniques