How to Cite:

Karwa, H., Gupta, P., Agrawal, R., Virdi, G. S., Kumar, A., & Jain, S. (2022). Multi-disease
prediction with machine learning. International Journal of Health Sciences, 6(S2), 9477–
9483. https://doi.org/10.53730/ijhs.v6nS2.7487

Multi-disease prediction with machine learning

Harsh Karwa
Student, CSE Department, Shri Ramdeobaba College of Engineering and
Management, Nagpur, India
Correspondence author email: [email protected]

Pavan Gupta
Student, CSE Department, Shri Ramdeobaba College of Engineering and
Management, Nagpur, India

Ram Agrawal
Student, CSE Department, Shri Ramdeobaba College of Engineering and
Management, Nagpur, India

Gursewak Singh Virdi

Student, CSE Department, Shri Ramdeobaba College of Engineering and
Management, Nagpur, India

Amit Kumar
Student, CSE Department, Shri Ramdeobaba College of Engineering and
Management, Nagpur, India

Sweta Jain
Professor, CSE Department, Shri Ramdeobaba College of Engineering and
Management, Nagpur, India

Abstract---In the present era, Machine learning (ML) algorithms are

extensively used in computer assisted diagnosis of the disease based
on the symptoms of the disease. The widespread use of healthcare
applications in the pandemic time, provides a motivation to further
develop new computer assisted diagnostic application in the
healthcare domain. Prevention and treatment of disease, accurate and
timely diagnosis of any health-related problem is essential. In the case
of a serious illness, a standard diagnostic method may not be enough.
We have proposed a system for predicting the disease. There were
about forty-one diseases in the data corpus that needed to be
analyzed based on the symptoms. The system delivers a disease
prediction that a person may have depending on the symptoms. This
diagnostic program can assist a physician in diagnosing disease,
allowing for timely treatment and saving lives. The disease forecasting
system was developed using ML models such as the Random Forests,

International Journal of Health Sciences ISSN 2550-6978 E-ISSN 2550-696X © 2022.

Manuscript submitted: 27 March 2022, Manuscript revised: 18 April 2022, Accepted for publication: 9 May 2022
9477
9478

the Naive Bayes, and the Support Vector Machine Classification

Algorithm. The presented work outlines an analysis of the
aforementioned algorithms.

Keywords---machine learning, disease prediction, random forest, naive

bayes, support vector machine.

Introduction

Medicine and health are important factors in economic growth and human life.
Technology assisted health care applications are significantly increasing since the
past two decades. In the pandemic period, there are many remote areas that still
do not have emergency health care services. To effectively cater the need of
masses in heavily populated. Country like India, the online diagnosis system is
the need of hour. Disease predicting systems may be a boom in many cases as it
can prevent the caused life risk diseases beforehand and can suggest the
individual to get immediate treatment to prevent further damage from the
underlying diseases. Not just that, this strategy will cut treatment costs and
reduce fear in the late stages, enabling sufficient treatment to be provided at the
right moment and reducing the rate of death. Furthermore, several localized
diseases have distinct features in different places, making disease outbreak
prediction difficult.

Related Work

There is a lot of research into disease prediction models utilizing various machine
learning algorithms, with varying results for various medical techniques. The
author Chauhan et al., (2020) shown an accuracy of the machine learning models
- Decision Tree, Random Forest, and Naive Bayes,as 92.4 %, 95.7 %, and 94.5 %
respectively[1]. Another study published by author Chen et al. (2017) on CNN-
based multimodal disease risk prediction achieved an accuracy of 94.5 % [2]. The
accuracy of the research work on Fuzzy Logic, Fuzzy Neural Networks, and
Decision Tree published by Leoni et al. (2017) was 58.8 %, 91 %, and 68.7 %,
respectively [3]. Furthermore, the accuracy achieved by Vijayarani et al. research
work on the SVM and Nayes Bayes was determined to be 79.66 % and 61.28 %,
respectively [4].

Materials and Methods

Our proposed work is based on the prediction of many diseases that follow a
patient's symptoms. In any medical application task the main important footstep
is to get the data corpus. The data preprocessing is an essential step to clean it
and prepare it for building a model during training phase. The testing of the
model is carried out using unseen data/test data from corpus.User will provide
the symptoms to our system. The symptoms will be provided as input / key
feature to our ML model where we will be using algorithms like Random Forest,
Naive Bayes, and SVM to predict disease in order to help the patient in early
stages of their disease. In this work, we have used python as a platform for using
 9479

machine learning algorithms. We've also built a great GUI to provide system
connectivity.

Dataset

The data corpus for this application is collected from Kaggle, which includes
attributes containing diseases, and their symptoms. The user needs to
understand related features in the dataset. This dataset can be easily found on
Kaggle for the same link that has been provided in references [5].Workflow
diagram of the system is shown in Figure1.

Dataset Collected from Kaggle

Preprocess Data and Split the dataset as

Training and Test Set

Build the ML Model using Training Set

RF, SVM, and NB

Test the model using Test Set

Predict the class of Disease

and evaluate the performance

Figure1: Workflow diagram of the system

Machine Learning

The ML algorithms are heavily dependent on the amount of the data presented to
them for learning / training the target variable. In this work we have implemented
multiclass classification of the diseases. The dataset is publically available.
Finding a subset of the relevant features is the major task in design of any
healthcare application. As, it is possible that many disease may have some of the
overlapping symptoms. The result obtained may not be generalized to real world
diagnosis applications, unless it is examined by the specialist doctor.

Naive Bayes

It's a classification technique that utilizes Bayes' Theorem as well as the predictor
independence condition. This model is straightforward and is extremely
9480

advantageous for big datasets. Naive Bayes is believed to exceed even the most
sophisticated classification algorithms due to its simplicity. This is rooted in the
Bayes theorem, which enables us to evaluate the conditional probabilities, say P
(F|G) using P (F), P (G), and P (G|F).Thus the Bayes Theorem can be represented
as

P (F|G) =P (G|F)*P (F)/P (G)

The conditional probabilities and the class probabilities P (Yi) are computed using
the training dataset by the Naive Bayes classifier. Although connected features
are voted twice in the model, the Naive Bayes classifier works perfectly when they
are omitted. This yields an overemphasis on the value of the associated features.

Random Forest

Random forest is a supervised learning technique that can be used to classify and
predict data. However, it is mostly employed to solve categorization issues. A
forest, as we all know, is made up of trees, and more trees equals a more healthy
forest. It's an ensemble method that's superior to a single decision tree because it
averages the results to reduce over fitting of the model. It chooses the best voting
solution. Random Forest produces better results than real problems mainly due
to noise incompatibility in the database and is not based on overload. It works
great too and shows excellent performance over other tree-based algorithms. To
read the tree, bootstrap is widely used for merging or wrapping.

Support Vector Machine

SVM is a popular method of classification. It is widely used in Machine Learning

for differentiating the differences in any given dataset. Linear SVM is used for
linear data, which means that if a database can be divided into two categories
using one straight line, such data is called linear data. Non-Linear SVM is used
for non-linear data, which means that if the database cannot be categorized using
a straight line. There are different types of kernels used in Non-Linear SVM, some
of them are Gaussian radial basis function (RBF), Polynomial kernel, Hyperbolic
Tangent kernel sigmoid kernel, ANOVA radial basis kernel. In our SVM, we have
utilized the RBF, which is the Gaussian radial basis function. The RBF kernel say
for a and a' can be represented as

Here,
= Square of the Euclidean distance between the two feature vectors
(For positive values)

Result and Discussion

This section indicates the results of a developed system that can predict disease
faster, accurately with high fidelity than the existing system. Results are obtained
 9481

with random forest, Naïve Bayes and SVM using Python. When a user accesses
the Disease Prediction Website, he or she is directed to the homepage. On the
homepage, there are specific procedures for predicting one's diseases, as seen in
figure 2.

The application also has a good visual interface that handles all of the inputs
required for prediction. The user will select symptoms from the drop down menu
and add them by clicking the add button; if the user wishes to remove a specific
symptom, he or she can do so by clicking the delete button and by clicking the
clear button, all symptoms are removed, as illustrated in figure 3.
9482

By clicking on predict now button, the user can find all of the probable diseases
predicted by the various algorithms, as seen in figure 4.

Conclusion

In this paper, we used three machine learning algorithms to predict and achieve a
desirable result for the user, as well as making the system more efficient than the
existing one and thus providing a better user experience than other available
systems. The present study focused only on the structured dataset of symptoms.
Most of the disease prediction system needs multimodal data as input for correct
diagnosis of the disease. There are no standard ways for dealing with semi-
structured and unstructured data.

References

1. “Disease Prediction using Machine Learning” Raj H. Chauhan, Daksh N.

Naik, Rinal A. Halpati, Sagarkumar J. Patel, Mr. A.D.Prajapati, International
Research Journal of Engineering and Technology (IRJET). Volume: 07 Issue:
05 | May 2020.
2. “Disease prediction by machine learning over big data from healthcare
communities”, M. Chen, Y. Hao, K. Hwang, L. Wang, IEEE Access, vol. 5, no.
1, pp. 8869-8879, 2017. DOI: 10.1109/ACCESS.2017.2694446
https://ieeexplore.ieee.org/abstract/document/7912315
3. “Disease Classification Using Machine Learning Algorithms - A Comparative
Study”, S. Leoni Sharmila, C. Dharuman and P. Venkatesan, , International
Journal of Pure and Applied Mathematics, vol. 114, no. 6, pp. 1-10, 2017.
4. “Liver Disease Prediction using SVM and Naive Bayes Algorithms” - Dr. S.
Vijayarani, Mr.S.Dayananda, International Journal of Science, Engineering
and Technology Research (IJSETR), 2015.Volume 4, Issue 4, April 2015.
 9483

5. Disease and symptomsDataset, Kaggle Dataset Link:

https://www.kaggle.com/itachi9604/disease-symptom-description-
dataset?select=dataset.csv