International Journal of Preventive Medicine and Health (IJPMH)
ISSN: 2582-7588 (Online), Volume-4 Issue-6, September 2024
Symptom Based Disease Prediction Using Machine
Learning
Ridham Sood, Virat Sharma
Abstract: The Disease Prediction Method utilises predictive And identifying various diseases. Predictive analysis using
modelling to forecast a user's disease based on the symptoms they practical multiple machine learning algorithms aids in more
provide as feedback to the system. Medical servicesare in dire need accurate disease prediction and treatment of patients.
of advancement to enable better decisions about patient care and
treatment options. In terms of machine learning, Healthcare Machine learning is now so pervasive that it is possible to use
enables humans to process large and complex medical databases, it many times a day without even realizing it. Machine
interpret them, and derive clinical insights. The machine analyses learning algorithms, on the other hand, only work with
the user's symptoms as input and returns the likelihood of the structured data and have a long computation time, as they
disease as output. Implementing the Decision Tree, K-Nearest store all the data in a training dataset and employ a complex
Neighbour, Naïve Bayes, and Random Forest enables disease calculation process.
prediction. In thispaper, we attempt to integrate machine learning
capabilities in healthcare into a single framework. Instead of
diagnosis, healthcare can be made smart by implementing disease II. PROBLEM DEFINITION
prediction using machine learning predictive algorithms. Whenan
The project's goal is to predict the disease by passing
early diagnosis of a disease is not possible, certain cases may arise.
As a result, disease prediction can be applied effectively. This symptomsinto it. Traditional disease risk models typically use
paper focuses primarily on creating a scheme, or what we would machine learning and supervised learning algorithms to train
call an immediate medical provision, that integrates symptoms the models, which utilise training data with labels.
obtained from multisensory devices, as well as other medical data, EHR records patient statistics, test results, and disease
and stores it in a healthcare dataset. This dataset will be analysed history, enabling the identification of potential data-driven
using a machine learning algorithm with an accuracy of more
strategies that reduce the cost of medical case studies. Bates
than 90%.
et al. propose six applications of big data in healthcare.
Keywords: Machine Learning Disease Prediction, Decision K Existing schemes can predict diseases, but not disease
Nearest Neighbor, Naïve Bayes, Random Forest.
subtypes. It is unable to predict people's health.
I. INTRODUCTION
III. PROBLEM SOLUTION
Machine learning is the process of programming The proposed system for disease prediction using machine
learning involves utilising various techniques, algorithms,
computers to improve their output based on previous data or
and tools to build a system that predicts a patient's disease
examples. The study of computer systems that learn from data
based on their symptoms. By comparing these symptoms with
and experienceis known as machine learning. There are two
the system’s previously available dataset, we can accurately
tracks in the machine learning algorithm: training and testing.
predict the disease. By taking those datasets and comparing
The prediction of a disease based on a patient's signs and
them with the patient’s disease, we will predict the accurate
medical history has become increasingly accurate with the
percentage of the disease in the patient.
advancement of machine learning over the past few decades.
The dataset and symptoms are passed to the system's
A portable health tracking system is proposed as a solution
prediction model, where the data is pre-processed for future
to the issue of advanced wearable devices. The primary
reference. Then, the user performs feature selection by
objective is to develop a wireless, low-cost, and user-friendly
entering various symptoms. Then, the classification of these
device that enables subjects to monitor clinical parameters
data is performed using different algorithms and techniques, such
such as body temperature and heart rate, allowing doctors to
as Decision Tree, KNN, Naïve Bayes, and Random Forest.
remotely and effectively manage the subject's disease.
The primary goal is to use machine learning in healthcare
IV. LITERARY SURVEY
to complement patient care and improve outcomes. Machine
learning has simplified the process of accurately diagnosing. A. Symptoms-Based Disease Prediction Using Decision
Treeand Electronic Health Record Analysis [1]
It has several attributes (symptoms) as well as classes
Manuscript received on 25 July 2024 | Revised Manuscript (diseases). To train the model, we use this to build the training
received on 13 August 2024 | Manuscript Accepted on 15
September 2024 | Manuscript published on 30 September 2024. and testing sets. We obtain the user's symptoms and use the
*Correspondence Author (s) qualified model to predict the disease. On the other side, the
Ridham Sood*, SRM Institute of Science and Technology, Chennai medical record is compiled, which results in a review of the
(Tamil Nadu), India. Email ID: rr8432@[Link] clinical report focusing on the most significant symptoms
Virat Sharma, SRM Institute of Science and Technology, Chennai (Tamil
Nadu), India. Email ID: vj6767@[Link] associated with a specific illness. This is used to increase the
number of disease symptom
© The Authors. Published by Lattice Science Publication (LSP). This is pairs in the dataset. This
an open access article under the CC-BY-NC-ND license
[Link]
paper describes disease
prediction using highly
Retrieval Number:100.1/ijpmh.G92340811922 Published By:
DOI: 10.54105/ijpmh.G9234.04060924 Lattice Science Publication (LSP)
Journal Website: [Link] 7 © Copyright: All rights reserved.
Symptom Based Disease Prediction Using Machine Learning
personalised training datasets, as well as related tasks such as higher accuracy.
scheduling appointments and locating the nearest healthcare
F. Review of Medical Disease Symptoms Prediction
centre.
Using Data Mining Technique [6]
B. Disease Prediction by Machine Learning Over Big The paper evaluates the performance of medical sickness
Data from Healthcare Communities [2] prediction supported by data processing techniques.
Using structured and unstructured data from hospitals, this Diagnosis of sicknesses like cancer, liver diseases, and heart
paper proposes a new convolutional neural network-based attacks was classified into numerous categories by the
multimodal disease risk prediction algorithm. To our classifier. By utilising two base classifiers, KNN and SVM,
knowledge, no current work in the field of medical big data the SVM technique achieved higher classification accuracy in
analytics has centred on both data forms. In comparison to compression compared to the standard cluster ensemble
other popular prediction algorithms, our proposed algorithm method. We tend to compare classification accuracy between
achieves a prediction accuracy of 94.8% and a convergence conventional algorithms and projected ensemble
speed that is fasterthan the CNN-based unimodal disease risk classification algorithms using the same dataset.
prediction algorithm. Additionally, to benchmark datasets from the UCI repository,
medical sickness datasets, and real-world datasets, we also
C. Multiple Disease Prediction Using Different
examined the classification accuracy.
Machine Learning Algorithms Comparatively [3]
This paper summarizes the findings of many studies in this G. Disease Prediction from Various Symptoms Using
area. Our proposed system aims to bridge the gap between Machine Learning [7]
doctors and patients, enabling all parties to achieve their The manuscript described a method of predicting disease
objectives. Using various Machine Learning algorithms, this based on a patient's symptoms, age, and gender. For
framework supports multiple disease predictions. Many systems' predicting diseases, the weighted KNN model performed
current approach focuses solely on automating this method, best, achieving an accuracy of 93.5% with the listed factors.
which falls short of establishing user confidence in the Almost all the machine learning models had good accuracy
system. By incorporating a doctor’s advice into our system, values. Due to the dependence of some models on parameters,
we maintain consumer trust while also ensuring that the the models did not provide accurate predictions, and their
doctor's business isnot harmed as a result of this system. accuracy rates were relatively low. Once the disease has been
predicted, we could easily manage the medicine resources
D. Multi-Disease Prediction using Data Mining required for treatment. As a result, the disease would be
Techniques [4] treated more affordably, and the recovery process would be
The accuracy of three data mining techniques is compared improved.
in this report. Along with high precision and recall metrics,
the aim is to provide high accuracy. While these metrics are V. ALGORITHMS
more commonly used in the field of information retrieval, A. Decision Tree
we've included them here because they're relevant to other
One of the most well-known machine learning algorithms is
metrics, including specificity and sensitivity. These metrics
the decision tree. A decision tree represents the decision logic
can be conveniently translated to true-positive (TP) and false-
forclassifying data items into a tree-like structure, i.e., tests
positive (FP) metrics using the uncertainty matrix. For
and outcomes. A decision tree's nodes usually have several
predicting various diseases, two separate data mining
levels, with the root node being the first or top-most node. All
classification strategies were employed, and their outputs
internal nodes (those with at least one child) are input
were compared to determine the most effective classifier.
variables or attribute checks. The classification algorithm
Building accurate andcomputationally effective classifiers for
branches towards the appropriate child node based on the test
medical applications isa significant challenge in data mining
result, andthe process of testing and branching repeats until
and machine learning.
the leaf nodeis reached.
E. Prediction of Heart Disease using Machine Learning
Algorithms [5]
What we discovered is that during small datasets and in
some other situations, decision trees often guide us to an
incorrect solution; however, when we look at Nave Bayes
results, we get more reliable results with probabilities for all
other possibilities, but decision trees can miss lead due to
guidance to only one solution. Finally, we can assume that
Naive Bayes is more reliable if the input data is clean and
well-maintained. While ID3 can clean itself, it cannot always
produce accurate results; similarly, Naive Bayes cannot
[Fig.1: Decision Tree]
always produce accurate results. We must consider the results
of various algorithms, and if a prediction is made using all of B. Naïve Bayes
their results, it will be correct. However, we can utilise Naive Bayes' theorem serves as
Bayes to consider variables individually and combine the foundation for the Naive
algorithms, such as Naive Bayes and K-means, to achieve Bayes classification
Retrieval Number:100.1/ijpmh.G92340811922 Published By:
DOI: 10.54105/ijpmh.G9234.04060924 Lattice Science Publication (LSP)
Journal Website: [Link] 8 © Copyright: All rights reserved.
International Journal of Preventive Medicine and Health (IJPMH)
ISSN: 2582-7588 (Online), Volume-4 Issue-6, September 2024
technique. This theorem can be used to explain the In a regression or classification query, it can be used to rank
probability of an occurrence based on prior knowledge of the values of variables naturally.
the event's conditions. This classifier assumes that a feature
in a class is not explicitly related to any other feature in the
class,even though features in that class can be interdependent.
The task of classifying a new entity into one of the existing
classes is used to demonstrate how the NB technique works.
[Fig.4: Random Forest]
VI. ARCHITECTURE OF SYMPTOM-BASED
DISEASE PREDICTION USING MACHINE
LEARNING
Fig. 2 Naïve Bayes
Machine learning-based disease prediction predicts the
C. K-nearest Neighbour
occurrence of a disease for a user based on different
One of the easiest and earliest classification algorithms is symptomsand the knowledge the user provides through the
the K-nearest neighbour algorithm. It's like a simplified symptoms. The architecture of the system for disease
version of a neural network classifier. The KNN algorithm,
prediction using machine learning consists of various
unlike the NB method, does not require the use of probability
datasets, from which we can compare and predict the user's
values. The number of nearest neighbours considered to take
symptoms. Afterwards, the datasets are transformed into
a ‘vote' is the ‘K' in the KNN algorithm. For the same sample
smaller sets and classified using classification algorithms.
item, different values for ‘K' may result in different
Finally, the classified data is processed using machine
classification results. Depicts the KNN's classification
learning technologies. The method then integrates and
process for a new object. When K=3, the new object (star) is
compares the above information and total processed data in
classified as ‘black', but when K=5, it is classified as ‘red’.
the prediction model, and ultimately predicts the disease.
[Fig.5: Architecture]
VII. RESULT
[Fig.3: K-Nearest Neighbour]
Higher precision can be achieved with the proposed
method. Based on the proposed algorithm, we not only use
D. Random Forest structured data but also the patient's text data. On the datasets,
we checkedour algorithms. On the dataset, we achieved 95%
A Random Forest is a set of classifiers based on Decision
accuracy for Decision Tree, Naive Bayes, and Random
Trees. A bootstrap sample of the data is used to construct
Forest, and 92.6% accuracy for K-Nearest Neighbour. The
each tree, which uses a candidate set of features chosen at
classification report mentioned below shows the accuracy of
random. For tree construction, it employs both bagging and
our algorithm for each disease, as well as the overall accuracy
random variable selection. After the forest has been
of our model.
developed, test cases are propagated down each tree, and the
trees make their classpredictions. A random forest's error rate
is determined by the intensity of each tree and the association
between any two trees.
Retrieval Number:100.1/ijpmh.G92340811922 Published By:
DOI: 10.54105/ijpmh.G9234.04060924 Lattice Science Publication (LSP)
Journal Website: [Link] 9 © Copyright: All rights reserved.
Symptom Based Disease Prediction Using Machine Learning
2. Chen, Yixue Hao, Kai Hwang, LuWang, Lin Wang p, (2017). Disease
Prediction Using Machine Learning on Big Data from Healthcare
Communities, Min IEEE. DOI:
[Link]
3. Rudra A. Godse, Smita S. Gunjal,Karan A. Jagtap, Neha S. Mahamuni,
Prof. Suchita (2019). Multiple Disease Prediction Using Different
Machine Learning Algorithms Comparatively Wankhade IJARCCE.
[Link]
content/uploads/2020/01/[Link]
4. [Link], Dr. D. Shanmuga Priyaa (2016). Multi-Disease Prediction
Using Data Mining Techniques, ResearchGate.
[Link]
rediction_using_Data_Mining_Techniques
5. Rajesh N, T Maneesha, Shaik Hafeez, Hari Krishna, (2018). Prediction of
Heart Disease Using Machine Learning Algorithms, ResearchGate. DOI:
[Link]
6. Rahul Deo Sah1, Dr. Jitendra Sheetalani (2017). Review of Medical
Disease Symptoms Prediction Using Data Mining Technique, IOSR-
JCE. DOI: [Link]
7. Rinkal Keniya, Aman Khakharia, Vruddhi Shah, Vrushabh Gada,
Ruchi Manjalkar, Tirth Thaker, Mahesh Warang, Ninad Mehendale
(2018). Disease prediction from various symptoms using machine
learning, SSRN. DOI: [Link]
Disclaimer/Publisher’s Note: The statements, opinions and
data contained in all publications are solely those of the
individual author(s) and contributor(s) and not of the Lattice
Science Publication (LSP)/ journal and/ or the editor(s). The
Lattice Science Publication (LSP)/ journal and/or the
editor(s) disclaim responsibility for any injury to people or
property resulting from any ideas, methods, instructions or
products referred to in the content.
[Fig.6: Classification Report]
VIII. CONCLUSION
An architecture diagram is a graphical representation of a
collection of concepts that comprise an architecture,
including itsvalues, elements, and components. The diagram
depicts the machine software in the context of a system
description.
DECLARATION STATEMENT
After aggregating input from all authors, I must verify the
accuracy of the following information as the article's author.
▪ Conflicts of Interest/Competing Interests: Based on my
understanding, this article does not have any conflicts of
interest.
▪ Funding Support: This article has not been sponsored or
funded by any organisation or agency. The independence
of this research is a crucial factor in affirming its
impartiality, as it was conducted without any external
influence.
▪ Ethical Approval and Consent to Participate: The data
provided in this article is exempt from the requirement for
ethical approval or participant consent.
▪ Data Access Statement and Material Availability: The
adequate resources of this article are publicly accessible.
▪ Author's Contributions: The authorship of this article is
contributed equally to all participating individuals.
REFERENCE
1. S Radhika, S Ramiya Shree, V Rukhmani Divyadharsini and A Ranjitha
(2020). Symptoms-Based Disease Prediction Using Decision Tree and
Electronic Health Record Analysis, European Journal of Molecular &
Clinical Medicine.
[Link]
[Link]
Retrieval Number:100.1/ijpmh.G92340811922 Published By:
DOI: 10.54105/ijpmh.G9234.04060924 Lattice Science Publication (LSP)
Journal Website: [Link] 10 © Copyright: All rights reserved.