Volume 7, Issue 11, November – 2022 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

Application of Machine Learning Techniques in

Prediction of Breast Cancer, Thyroid, and PCOS:
An Overview
Akanksha Gautam, Sunita Jalal, Chetan Singh Negi
Department of Computer Engineering, College of Technology, GBPUAT Pantnagar

Abstract:- Because of the rapid population growth in disorder. Malignant cells evolve destructively and uncontrolled
diseases in recent years, early disease identification has way. Breast cancer symptoms include lumps in the breast, shape
become a critical problem. With massive population changes, skin dimpling, and flaky patches. Changes in eating
growth, the probability of death from breast cancer is habits, a lack of fitness, chemical substitution therapy during
increasing exponentially. Breast Cancer is the second the menopausal phase, and consistent family history are all
leading most severe of the cancers that have already been significant causes of these manifestations. According to
identified, and following this, others are Thyroid and estimates, approximately 316,120 females were diagnosed with
Polycystic Ovary Syndrome (PCOS) disease. An automatic various breast cancers in the USA in 2017.
disease diagnosis system assists medical personnel in disease
detection, provides dependable, effective, and immediate  Thyroid:
responses, and reduces the death risk. In this paper, we Thyroid disease is one of the most common public health
study different machine learning techniques and their issues, particularly among pregnant women. It harms
utility in predicting breast cancer, thyroid and PCOS to get metabolism, mental functions, energy level, weakness, and
the best result. The performance measures such as bowel movements. Most thyroid diseases diagnosed during
accuracy, specificity, sensitivity, precision, recall, F1 score, pregnancy are caused by a lack or excess of iodine in the diet.
and receiver operating curve are discussed to assess the Thyroid disease poses a significant challenge to pregnant
performance of machine learning algorithms. The paper women's physiology and has substantial maternal and foetal
explores the research work using machine learning implications. There is a growth in thyroxine (T4) and
algorithms done in detecting breast cancer, thyroid, and triiodothyronine (T3) generation in women with normal thyroid
PCOS. function. It hampers thyroid-stimulating hormone (TSH) in the
first three months of pregnancy due to an increased human
Keywords:- Breast Cancer, Thyroid, PCOS, Dataset, Machine chorionic gonadotropin (hCG) level, which enables the TSH
Learning. receptor.

I. INTRODUCTION  PCOS:
PCOS is a disease in which the ovary creates a varying
Women's health varies from men's in numerous ways. amount of androgens hormone, generally present in small
Women's health is an illustration of population health, which amounts in women. During the reproductive period,
the World Health Organization defines as "a state of complete approximately 6-10% of women develop Polycystic Ovary
physical, mental, and social well-being, rather than merely the Syndrome (PCOS). It manifests as various symptoms in women
absence of disease or infirmity." While the accelerations of the of reproductive age resulting from a hormonal disorder. PCOS
foremost causes of death, cancer, cardiovascular and lung manifests itself in three ways. The first is a menstrual
diseases, are comparable in men and women, women have irregularity, followed by an increase in male hormones in
distinctive experiences. Lung cancer has surpassed all other women, and finally, an ultrasound-detected polycystic ovarian
cancers as the foremost reason for cancer death in females, structure. Polycystic Ovary Syndrome is suspected in women
followed by various cancers such as breast cancer, ovarian who exhibit these symptoms, but the diagnosis is determined by
cancer, colorectal cancer, uterine cancer, and cervical cancer. examining ultrasound images and blood test results.

 Breast Cancer: Machine learning (ML) refers to an algorithm's ability to

Today, breast cancer is a common and dangerous 'learn' by detecting patterns in large datasets. In other words, the
condition. According to World Health Organization research, 'outcomes' yielded by machine learning algorithms are
1,000,000 women are practically sure to acquire breast cancer, statistical inferences drawn from huge datasets. Machine
and half will die because the disease is often detected late. One learning techniques utilise vast amounts of computing power to
in every eight women in the US will be diagnosed with breast determine rules the human brain would be unable to process.
cancer. Breast cancer is the most dominant breast malignant The more data fed into a machine learning model, the more

IJISRT22NOV1357 www.ijisrt.com 1823

Volume 7, Issue 11, November – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
complex the rules become – and the more exact the predictions Table 1: Details of Attributes
become. It is an artificial intelligence application in which Name Type
computers are programmed to mimic how humans think and Character of background Nominal - F: Fatty, G:Fatty-
learn. Machine learning helps in healthcare collect and manage tissue glandular, D:Dense-glandular
patient data, identify trends, recommend treatments, and more. Classes of Abnormality Nominal- CALC, CIRC, SPIC,
Hospitals and healthcare organisations have begun recognising MISC, ARCH, ASYM,
machine learning's ability to improve decision-making and NORM
reduce risk in the medical field, resulting in a slew of new and Severity of Abnormality Nominal: Benign, Malignant
exciting career opportunities. Machine learning aims to x image coordinate of Integer
enhance patient outcomes and generate previously unavailable centre of abnormality
medical insights. Jalal et al. [22] discussed the use of machine y image coordinate of Integer
learning techniques: Latent Dirichlet Allocation and k-Medoids centre of abnormality
in web services clustering. Web services can provide access to Approximate radius of a pixels
different modules of healthcare application software. Arushi et circle enclosing the
al. [23] explored the application of various machine learning abnormality
models in Covid-19 prediction. The paper is organized as
follows. Section 2 gives descriptions of breast cancer, thyroid, B. Thyroid Dataset:
and PCOS datasets available in different repositories. The The different thyroid disease data sets are available on
overview of various machine learning algorithms and UCI machine learning repository. These source datasets are
performance measures is presented in section 3. Section 4 classified using different machine learning algorithms by
outlines the related work in detecting breast cancer, thyroid, and various researchers.
PCOS. The conclusion of the paper is drawn in section 5.
Throid dataset [27]: The dataset contains 7200 instances
II. DATASETS DESCRIPTIONS and 29 attributes. Some important attributes are age, sex,
Thyroid stimulating hormone (TSH), triiodothyronine(T3),
This section gives an overview of different dataset used Thyroid binding globulin(TBG), Thyroxin utilization
by researchers in detection of breast cancer, thyroid, PCOS. rate(T4U), Total Thyroxin, Free(TT4) Thyroxin Index (FTI).
A. Breast Cancer Datasets: C. PCOS Dataset:
Wisconsin Breast Cancer Dataset1 [24]: The dataset is Machine learning algorithms are used to diagnose
available on the UCI repository, and it is of the year 1992. It polycystic ovary syndrome (PCOS) in women using a dataset
contains 699 instances where the cases are either benign or freely available in the Kaggle repository, the dataset is called
malignant. There are 458 benign cases (65.50%) and 241 Polycystic Ovary Syndrome.
malignant cases (34.50%). The dataset has two classes such as
2 or 4, where 2 represents the benign case, and 4 illustrates the PCOS dataset [28]: The dataset is available on github. It
malignant case. The dataset contains the following attributes: contains 119 instances with 18 attributes. Some of the attributes
Clump Thickness, Uniformity of Cell Size, Uniformity of Cell are: Regularity of menstrual periods, weight gain, depression
Shape, and Marginal Adhesion. and anxiety, history of diabetes and hyper tension, fast food
intake.
Wisconsin Breast Cancer Dataset2 [25]: The dataset is
published in the year 1995. It is available on the kaggle PCOS dataset [29]: The dataset is available on Kaggle
repository. The dataset contains 569 samples. The dataset is repository. It contains 541 instances with 43 attributes. Clinical
popularly used for many applications because it includes noise- parameters are TSH, FSH, Hb, AMH, LH Vit, D3, PRG, and
free instances. The dataset considers ten attributes: texture, other metabolic, hormonal, and biochemical indicators.
radius, smoothness, area, perimeter, concave points, symmetry, Whereas physical factors included measurements of height,
compactness, concavity, and fractal dimension. Attributes are weight, age, waist, and hip.
real-valued.
III. OVERVIEW OF MACHINE LEARNING
MIAS Dataset [26]: Some researchers used MIAS dataset. TECHNIQUES AND EVALUATION MEASURES
They used machine learning techniques on dataset to detect
breast cancer in mammograms. The dataset contains 322 This section explores different machine learning
sample instances. In which nearly 189 images were of normal techniques such as Logistic Regression, Linear Regression,
cases and 133 images were of abnormality. The attributes Naive Bayes classification technique, K-Nearest Neighbors
presented in the dataset are defined in Table 1: (KNN), Decision Tree Random Forests, Artificial Neural
Networks (ANNs), Support Vector Machine (SVM), and deep
learning. Some evaluation measures such as accuracy,

IJISRT22NOV1357 www.ijisrt.com 1824

Volume 7, Issue 11, November – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
sensitivity, specificity, precision, recall, and F1 score are of a single classification tree, RF generates a forest of
discussed. classification trees from a given dataset.

A. Machine Learning Techniques  K-Nearest Neighbors:

The supervised classification algorithm k-Nearest
 Logistic Regression: Neighbors (K-NN) takes a large number of categorized points
Logistic Regression is an analytical modeling technique and uses them to show itself how to tag other points. The K-
that associates a level's likelihood with a set of explicative nearest neighbor algorithm is used for pattern recognition and
variables. It analyses a dataset in which one or more grouping. It is widely used in forecasting. When new data
independent variables determine the outcome. A binary arrives, the K-NN algorithm identifies existing data points that
variable is used to assess the outcome (in which there are only are closest to it. Any attribute that varies on a large scale may
two possible results). Given a set of independent variables, it have a significant impact on the interval between data points.
predicts a binary result (True/False, 1/0, Yes/No). K-NNs are based on the assumption that data samples are
represented in a metric space.
 Linear Regression:
The linear regression algorithm shows a linear  Support Vector Machine (SVM):
relationship between a dependent (y) variable and one or more The support vector machine is an extension of a natural
independent (x) variables. Because linear regression classifier known as the maximal edge classifier. The meaning
demonstrates a linear relationship, it determines how the value of hyperplane in an n-dimensional space is accompanied by the
of the dependent variable changes with the value of the maximal edge classifier. The hyperplane has (n- 1) dimensions
independent variable. Different regression models differ in the and a level subspace that does not have to go through the root.
type of relationship they consider between dependent and Because drawing a hyperplane in a higher dimension is
independent variables, as well as the number of independent difficult, (n- 1) dimensional level subspace is still used. In p-
variables used. dimensions, the hyperplane used is as follows:

 Naive Bayes Technique: 𝛽 0 + 𝛽1X1 + 𝛽2X2 + ⋯ + 𝛽p Xp = 0

The Nave-Bayes method is a supervised learning
algorithm that customs the Bayes theorem to resolve where X1, X2,..., and Xp are data points in the p-
classification difficulties. It is a simple and effective classifier dimensional sample space and 0, 1, 2,..., and p are hypothetical
primarily used for text classification with an extensive training values.
set. The Nave Bayes Classifier helps develop fast machine
learning models capable of constructing quick predictions. It  Artificial Neural Networks (ANNS):
proposes classification based on an object's probability. The artificial neural network algorithm is inspired by
biological neurons and works by mimicking the dendrite, soma,
 Decision Tree: and axon workflow. An artificial neural network's basic
The decision tree classification method classifies training architecture is a collection of interconnected neurons organized
data into a tree. We can find rules from the tree. The into three layers: input, hidden, and output. In general, this type
performance of the decision tree is determined using test data of network learns to perform tasks by considering a sufficient
drawn randomly from training data using the hold-out method. number of examples.

 Random Forests:  Deep Learning:

Random Forest classifier utilises many decision trees on The deep learning technique teaches computer machines
various subsets of a dataset and computes average accuracy to to do what human beings instinctively do: learn by doing. Deep
enhance the model's predictive accuracy. Rather than learning is the leading technology behind self-driving cars,
depending on the decision of a single decision tree, the random letting them understand a stop sign or differentiate between a
forest considers the assumptions from each tree, and it road user and a physical static object. Deep learning allows a
computes the final outcome based on the majority of votes computer-based system to perform classification tasks directly
concept. The considerable number of trees increases the from images, text, or sound.
accuracy and lowers the risk of overfitting. The random forest B. Evaluation Measures
classifier is an extremely effective supervised classification
tool. The RF classification is an ensemble method that can be  Confusion Matrix:
thought of as a type of nearest neighbour predictor. Ensemble A confusion matrix contains summarized data of
learning is the process of strategically developing and prediction used to present the model's performance. It is a table
incorporating statistical methods such as classifiers or experts that contains both predicted and actual values.
to solve a specific computational intelligence problem. Instead

IJISRT22NOV1357 www.ijisrt.com 1825

Volume 7, Issue 11, November – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Actual\Predicted A B IV. LITERATURE REVIEW
A TP FN
B FP TN This section presents existing research works done in
prediction of breast cancer, thyroid, and PCOS.
Here true positive TP is the outcome that means machine
A. Machine learning techniques used in prediction of breast
learning model predicts Class A correctly. True negative value
cancer:
TN is the outcome that the model accurately predicts Class B.
Gopal et al. [1] explored the machine learning techniques:
False positive FP denotes the result that the classification model
Naïve Bayes, linear regression, logistic regression, SVM,
mispredicts Class B. False negative value FN defines the
decision tree, random forest, and neural network in predicting
outcome that the model mispredicts Class A.
breast cancer with IOT devices. The main objective of the
research work done in [2] is to predict and diagnosis breast
 Accuracy:
cancer, using machine-learning algorithms: Logistic
Accuracy of a machine learning model gives the
Regression ,Support Vector Machine (SVM), Random Forest
percentage of tuples correctly classified by the model.
,K-Nearest Neighbors (KNN) and Decision tree (C4.5), and
𝑎𝑐𝑐𝑢𝑟𝑎𝑐𝑦 find out the most effective results with respect to confusion
𝑛𝑜. 𝑜𝑓 𝑡𝑟𝑢𝑒 𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒 + 𝑛𝑜. 𝑜𝑓 𝑡𝑟𝑢𝑒 𝑛𝑒𝑔𝑎𝑡𝑖𝑣𝑒
= matrix, accuracy and precision. In this study [3], the authors
(𝑛𝑜. 𝑜𝑓 𝑡𝑟𝑢𝑒 𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒 + 𝑛𝑜. 𝑜𝑓 𝑡𝑟𝑢𝑒 𝑛𝑒𝑔𝑎𝑡𝑖𝑣𝑒 + 𝑛𝑜. 𝑜𝑓 𝑓𝑎𝑙𝑠𝑒 𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒 + 𝑛𝑜. 𝑜𝑓 𝑓𝑎𝑙𝑠𝑒 𝑛𝑒𝑔𝑎𝑡𝑖𝑣𝑒)
described a Computer-Aided Diagnosis (CAD) technique with
deep learning to classify breast cancer patients into 3 classes
 Sensitivity or Recall: (cancer, no cancer, and non-cancerous). Islam et al. [4]
Sensitivity is simply the true positive rate. It can also be presented a model that predicts breast cancer disease using
defined as the proportion of correctly classified positive tuples. machine learning techniques such as SVM and K-Nearest
Neighbors. The SVM-based model works with an accuracy
𝑆𝑒𝑛𝑠𝑖𝑡𝑖𝑣𝑖𝑡𝑦 𝑜𝑟 𝑅𝑒𝑐𝑎𝑙𝑙 value of 99.68% in the training phase. The proposed model is
𝑛𝑜. 𝑜𝑓 𝑡𝑟𝑢𝑒 𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒 beneficial for the patient and medical team. Yue et al. [5]
= reviewed machine learning techniques such as artificial neural
(𝑛𝑜. 𝑜𝑓 𝑡𝑟𝑢𝑒 𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒 + 𝑛𝑜. 𝑜𝑓 𝑓𝑎𝑙𝑠𝑒 𝑛𝑒𝑔𝑎𝑡𝑖𝑣𝑒) network, support vector machine, k-nearest neighbor, decision
tree and their applications in breast cancer diagnosis and
 Specificity:
prediction. Gayathri et al. [6] summarized the works done on
Specificity is defined as negative rate that is the fraction breast cancer diagnosis using different machine learning
of correctly classified negative tuples. methods to enhance the accuracy of detecting breast cancer.
The study helps us understand the research works proposed to
𝑆𝑝𝑒𝑐𝑖𝑓𝑖𝑐𝑖𝑡𝑦 diagnose breast cancer.
𝑛𝑜. 𝑜𝑓 𝑡𝑟𝑢𝑒 𝑛𝑒𝑔𝑎𝑡𝑖𝑣𝑒
=
(𝑛𝑜. 𝑜𝑓 𝑓𝑎𝑙𝑠𝑒 𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒 + 𝑛𝑜. 𝑜𝑓 𝑡𝑟𝑢𝑒 𝑛𝑒𝑔𝑎𝑡𝑖𝑣𝑒) Islam et al. [7], compared the use of K-nearest neighbours,
support vector machine (SVM), logistic regression, artificial
 Precision: neural networks (ANNs), and random forests supervised
It calculates the proportion of true positive versus all machine learning techniques in breast cancer prediction. The
positive outcomes. sensitivity, accuracy, precision, F1 score, negative predictive
value, specificity, Matthews Correlation Coefficient, false-
𝑃𝑟𝑒𝑐𝑖𝑠𝑖𝑜𝑛 negative rate, and false-positive rate are used to assess the
𝑛𝑜. 𝑜𝑓 𝑡𝑟𝑢𝑒 𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒 performance of machine learning techniques. Kumar et al. [8]
=
(𝑛𝑜. 𝑜𝑓 𝑓𝑎𝑙𝑠𝑒 𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒 + 𝑛𝑜. 𝑜𝑓 𝑡𝑟𝑢𝑒 𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒) explored and compared the performance of J48, Naïve Bayes,
and SVM classification algorithms and combination of these
 F1 Score: algorithms using voting classifier technique for the prediction
F1 score is computed by taking harmonic mean of of breast cancer. Voting is one of the ensemble approaches
precision and recall. where we can combine multiple models for the better
classification. In [9], CART classifier has been applied for the
2 ∗ (𝑃𝑟𝑒𝑐𝑖𝑠𝑖𝑜𝑛 ∗ 𝑅𝑒𝑐𝑎𝑙𝑙) classification of various breast cancer datasets. The bagging
𝐹1 𝑠𝑐𝑜𝑟𝑒 =
(𝑃𝑟𝑒𝑐𝑖𝑠𝑖𝑜𝑛 + 𝑅𝑒𝑐𝑎𝑙𝑙) technique is used to assess the classifier's performance
concerning accuracy and classification time. Harinishree et al.
 Receiver Operating Curve: [10] examined different datasets and machine learning models
It is a graphical aid to compare the performance of two or and accuracy of various types of machine leaning models used
more machine learning models. It plots true positive rate against to predict breast cancer. Yarabarla et al. [11] discussed the use
false positive rate. of KNN and random forest machine learning models to predict
whether the patient is having breast cancer or not.

IJISRT22NOV1357 www.ijisrt.com 1826

Volume 7, Issue 11, November – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
B. Machine learning techniques used in prediction of thyroid: C. Machine learning techniques used in prediction of PCOS
Raisinghani et al. [12] discussed the most recent machine Denny et al. [17] developed a method for the initial finding
learning algorithms: Support vector machine, Decision trees, and estimate of PCOS focused on an optimal and slight but
Logistic regression, Random forest, and Multilayer Feed promising set of medical and metabolic limitations that act as
Forward Neural Network used in the detection and prediction an early marker for the PCOS disease. It primarily focuses on
of thyroid detection. The proposed scheme is used to predict creating algorithms that can access the datasets offered and
thyroid disease among patients based on various thyroid utilize the data for the network's learning purposes. Bharti et al.
symptoms and reports. Duggal et al. [13],discussed several [18] explained its model-based detection for polycystic ovary
feature extraction and classification techniques for thyroid syndrome in women patients. Machine learning algorithms
disease diagnosis which are related to machine learning such as logistic regression, gradient boosting, random forest,
techniques problems. The methods for selecting features and hybrid random forest and logistic regression (RFLR) are
introduced are Tree Based Feature Selection, Recursive Feature used to apply to a dataset that is openly accessible in the Kaggle
Elimination, and, Univariate Selection. Three classification source. Chauhan et al. [19] used Machine Learning techniques
techniques were used: Support Vector Machines, Random such as Decision Tree Classifier, K-Nearest Neighbor(KNN),
Forest, and Naive Bayes. Abbad et al. [14], examined various Logistic Regression(LR), Naive Bayes, Support Vector
classifiers such as logistic regression, Naive Bayes, decision Machine(SVM) to create an application for initial PCOS guess.
tree, K-nearest neighbour (KNN), and support vector machine The Decision Tree Classifier was discovered to be the most
that are implemented with or without feature selection adequate model for PCOS forecast based on the accuracy and
techniques. Tyagi et al. [15], introduced various machine confusion matrix. An application was created to assist users in
learning techniques such as K-NN, support vector machine, predicting PCOS at an early stage. Vikas et al. [20] compared
decision tree and analysis for thyroid preventative measure to the accuracies and other concert methods of the previously
predict a patient's risk of developing thyroid disease. Anari et mentioned data mining techniques such as Neural Networks,
al. [16] explained the goals of deep learning in thyroid cancer Decision Tree and Naïve Bayes for predicting whether a patient
imaging and performed a review of literature on its limitations, is likely to have PCOS or not. In [21], seven classifiers were
potential, and current application in this field. They provided an used to apply various machine algorithms. The results show that
overview of recent developments in thyroid cancer analysis the Linear Discriminant classifier performs best in terms of
using deep learning methods and mentioned different accuracy, while the KNN classifier performs best in terms of
challenges and practical issues that may limit the growth and sensitivity. Table 2 compares research works done in predicting
integration of deep learning into medical workflow. breast cancer, thyroid, and PCOS using different machine
learning techniques.

Table 2: Comparative Analysis of Existing Works

Reference Disease Name Machine Learning Techniques Dataset Evaluation Measures
[1] Breast Cancer Logistic regression, Naïve Bayes, SVM, Wisconsin Prognostic Precision, Recall, F-
linear regression, random forest, decision Breast Cancer Measure and accuracy
tree, neural network Chemotherapy
(WPBCC), Wisconsin
Diagnostic Breast
Cancer (WDBC)
[2] Breast Cancer Support Vector Machine (SVM), K- Wisconsin Breast Accuracy
Nearest Neighbors (KNN), Random Cancer Diagnostic
Forest, Logistic Regression, and Decision
tree (C4.5)
[3] Breast Cancer Support Vector Machine (SVM), Deep Dataset from Kaggle Accuracy
Learning- Convolutional Neural Network repository
(CNN), and Random Forest
[4] Breast Cancer K-Nearest Neighbors and Support Vector Wisconsin breast Accuracy, False
Machine cancer diagnosis data Discovery Rate
Specificity, Sensitivity,
[5] Breast Cancer Artificial neural network, SVM, decision Wisconsin breast Accuracy
trees, and K-Nearest Neighbors. cancer database
(WBCD)
[6] Breast Cancer Reinforcement learning, Supervised Wisconsin Breast Accuracy
learning, Unsupervised learning, Semi- Cancer Diagnostic
supervised learning, Transduction,
Learning to learn.

IJISRT22NOV1357 www.ijisrt.com 1827

Volume 7, Issue 11, November – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
[7] Breast Cancer Random forests, Support vector machine Wisconsin Breast Accuracy,
(SVM), logistic regression, artificial Cancer Precision,F1score,
neural networks (ANNs) and K-nearest
neighbors
[8] Breast Cancer J48, Naïve Bayes, and SVM Wisconsin University Accuracy, Sensitivity,
database Specificity and
misclassification rate
[9] Breast Cancer Decision tree, bagging, CART Wisconsin University Accuracy
database
[10] Breast Cancer Supervised Learning, Neural Networks, Wisconsin Breast Accuracy
Naïve Bayes, Decision Tree Cancer Dataset, MIAS
Dataset
[11] Breast Cancer KNN, Random forest WBCD data set from Recall,F1-score,support
UCI machine learning
repository
[12] Thyroid Decision trees, Random forest, Support UCI machine learning Accuracy, Precision, F1
vector machine, Logistic regression and repository score, Recall
Multilayer Feed Forward Neural Network
[13] Thyroid Support vector machines, Naive Bayes, UCI Machine Learning Accuracy
and Random Forest Repository
[14] Thyroid L1 and L2 Norm‑Based Model Feature DHQ Teaching Precision, Accuracy, and
Selection, KNN, SVM, Naive Bayes, Hospital, Dera Ghazi Receiver Operating
Decision Tree and Logistic regression Khan, Pakistan Curve
[15] Thyroid Support vector machine (SVM), K-NN, UCI machine learning Accuracy, Mean absolute
Decision Trees repository error
[16] Thyroid Deep learning, ANN, CNN, GAN, Accuracy, Sensitivity,
- Specificity
[17] PCOS Logistic Regression, Naïve Bayes Sampled datasets of Accuracy, Sensitivity,
Classifier, K-NN, Classification and 541 samples Specificity, Precision, F1
Regression Trees (CART), Support Vector score
Machine (SVM), Random Forest
Classifier,
[18] PCOS Gradient boosting, random forest, logistic Dataset from Kaggle Accuracy
regression, and hybrid random forest and repository
logistic regression (RFLR)
[19] PCOS Logistic Regression, K-Nearest Neighbor, Sampled datasets of Accuracy, Precision,
Naive Bayes, Decision Tree, Support 267 samples Recall, Specificity, F1
Vector Machine, score
[20] PCOS Naïve Bayes, Decision Tree and Neural Survey conducted Sensitivity, Specificity,
Networks based on the lifestyle of Precision, Accuracy
the women.
[21] PCOS Logistic Regression, Linear Regression, Kaggle Accuracy, Precision,
KNN, Naïve Bayes, SVM, Decision Tree, dataset name Polycystic Sensitivity, Specificity
Ovary Syndrome.

V. CONCLUSION recall, f1 score, and receiver operating curve. The study

presented in the survey concludes that machine learning
This survey presents an overview of applications of techniques provide automatic disease detection with high
different machine learning techniques in predicting breast performance.
cancer, thyroid and PCOS. Existing research works incorporate
various machine learning algorithms such as Support vector REFERENCES
machine (SVM), K-nearest neighbors, decision trees, random
forests, artificial neural networks (ANNs), logistic regression, [1]. Gopal, V. N., Al-Turjman, F., Kumar, R., Anand, L., &
Naive Bayes techniques, and deep learning. These machine Rajesh, M.(2021).Feature selection and classification in
learning models are evaluated using different performance breast cancer prediction using IoT and machine
measures such as accuracy, sensitivity, specificity, precision, learning. Measurement,178, 109442.

IJISRT22NOV1357 www.ijisrt.com 1828

Volume 7, Issue 11, November – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
[2]. Naji, M. A., El Filali, S., Aarika, K., Benlahmar, E. H., algorithms for thyroid disease. Arabian Journal for
Abdelouhahid, R. A., & Debauche, O. (2021). Machine Science and Engineering, 46(10), 9437-9449.
learning algorithms for breast cancer prediction and [15]. Tyagi, A., Mehra, R., & Saxena, A. (2018, December).
diagnosis. Procedia Computer Science, 191, 487-492. Interactive thyroid disease prediction system using
[3]. Allugunti, V. R. (2022). Breast cancer detection based on machine learning technique. In 2018 Fifth international
thermographic images using machine learning and deep conference on parallel, distributed and grid computing
learning algorithms. International Journal of Engineering (PDGC) (pp. 689-693). IEEE.
in Computer Science, 4(1), 49-56. [16]. Anari, S., Tataei Sarshar, N., Mahjoori, N., Dorosti, S., &
[4]. Islam, M. M., Iqbal, H., Haque, M. R., & Hasan, M. K. Rezaie, A. (2022). Review of Deep Learning Approaches
(2017, December). Prediction of breast cancer using for Thyroid Cancer Diagnosis. Mathematical Problems in
support vector machine and K-Nearest neighbors. In 2017 Engineering, 2022.
IEEE Region 10 Humanitarian Technology Conference [17]. Denny, A., Raj, A., Ashok, A., Ram, C. M., & George, R.
(R10-HTC) (pp. 226-229). IEEE. (2019, October). i-hope: Detection and prediction system
[5]. Yue, W., Wang, Z., Chen, H., Payne, A., & Liu, X. (2018). for polycystic ovary syndrome (pcos) using machine
Machine learning with applications in breast cancer learning techniques. In TENCON 2019-2019 IEEE Region
diagnosis and prognosis. Designs, 2(2), 13. 10 Conference (TENCON) (pp. 673-678). IEEE.
[6]. Gayathri, B. M., Sumathi, C. P., & Santhanam, T. (2013). [18]. Bharati, S., Podder, P., & Mondal, M. R. H. (2020, June).
Breast cancer diagnosis using machine learning Diagnosis of polycystic ovary syndrome using machine
algorithms–a survey. learning algorithms. In 2020 IEEE Region 10 Symposium
[7]. Islam, M., Haque, M., Iqbal, H., Hasan, M., Hasan, M., & (TENSYMP) (pp. 1486-1489). IEEE.
Kabir, M. N. (2020). Breast cancer prediction: a [19]. Chauhan, P., Patil, P., Rane, N., Raundale, P., & Kanakia,
comparative study using machine learning techniques. SN H. (2021, June). Comparative analysis of machine
Computer Science, 1(5), 1-14. learning algorithms for prediction of pcos. In 2021
[8]. Kumar, U. K., Nikhil, M. S., & Sumangali, K. (2017, International Conference on Communication information
August). Prediction of breast cancer using voting classifier and Computing Technology (ICCICT) (pp. 1-7). IEEE.
technique. In 2017 IEEE international conference on [20]. Vikas, B., Anuhya, B. S., Chilla, M., & Sarangi, S. (2018).
smart technologies and management for computing, A critical study of Polycystic Ovarian Syndrome (PCOS)
communication, controls, energy and materials classification techniques. IJCEM Int J Comput Eng
(ICSTM) (pp. 108-114). IEEE. Manage, 21(4).
[9]. Lavanya, D., & Rani, K. U. (2012). Ensemble decision [21]. Hdaib, D., Almajali, N., Alquran, H., Mustafa, W. A., Al-
tree classifier for breast cancer data. International Journal Azzawi, W., & Alkhayyat, A. (2022, May). Detection of
of Information Technology Convergence and Polycystic Ovary Syndrome (PCOS) Using Machine
Services, 2(1), 17-24. Learning Algorithms. In 2022 5th International
[10]. Harinishree, M. S., Aditya, C. R., & Sachin, D. N. (2021, Conference on Engineering Technology and its
April). Detection of Breast Cancer using Machine Applications (IICETA) (pp. 532-536). IEEE.
Learning Algorithms–A Survey. In 2021 5th International [22]. Jalal, S., Yadav, D. K., & Negi, C. S. (2019). Web service
Conference on Computing Methodologies and discovery with incorporation of web services
Communication (ICCMC) (pp. 1598-1601). IEEE. clustering. International Journal of Computers and
[11]. Yarabarla, M. S., Ravi, L. K., & Sivasangari, A. (2019, Applications, 1-12.(Online)
April). Breast cancer prediction via machine learning. In [23]. Arushi , Jalal, S. , Negi, C. S. (2022, October). Covid-19
2019 3rd International Conference on Trends in Prediction based on Symptoms using Machine Learning,
Electronics and Informatics (ICOEI) (pp. 121-124). INTERNATIONAL JOURNAL OF ENGINEERING
IEEE. RESEARCH & TECHNOLOGY (IJERT), 11(10).
[12]. Raisinghani, S., Shamdasani, R., Motwani, M., Bahreja, [24]. Breast Cancer Wisconsin (Original) Dataset, [Online].
A., & Raghavan Nair Lalitha, P. (2019, April). Thyroid https://archive.ics.uci.edu/ml/machine-learning-
prediction using machine learning techniques. databases/breast-cancer-wiscon sin/breast-cancer-
In International Conference on Advances in Computing wisconsin.data
and Data Sciences (pp. 140-150). Springer, Singapore. [25]. Breast Cancer Wisconsin dataset,[Online].
[13]. Duggal, P., & Shukla, S. (2020, January). Prediction of https://www.kaggle.com/datasets/uciml/breast-cancer-
thyroid disorders using advanced machine learning wisconsin-data
techniques. In 2020 10th International Conference on [26]. Breast Cancer Dataset, [Online].
Cloud Computing, Data Science & Engineering http://peipa.essex.ac.uk/info/mias.html
(Confluence) (pp. 670-675). IEEE. [27]. Thyroid dataset. http://archive.ics.uci.edu/ml/machine-
[14]. Abbad Ur Rehman, H., Lin, C. Y., Mushtaq, Z., & Su, S. learning-databases/thyroid-disease/
F. (2021). Performance analysis of machine learning

IJISRT22NOV1357 www.ijisrt.com 1829

Volume 7, Issue 11, November – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
[28]. PCOS dataset. https://github.com/PCOS-
Survey/PCOSData/blob/master/PCOS-Data.csv
[29]. PCOS dataset.
https://www.kaggle.com/datasets/prasoonkottarathil/poly
cystic-ovary-syndrome-pcos

IJISRT22NOV1357 www.ijisrt.com 1830