Volume 9, Issue 5, May – 2024 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAY1274

Diabetes Prediction System Using SVM Alogrithm

Snehal Mhatre1; Harshada Dixit2; Snehal Jagdale3; Shital Narsale4; Naufil Kazi5
Information Technology Bharati Vidyapeeth DU Engineering
and Technology, Navi Mumbai, Navi Mumbai, India

Abstract:- Diabetes Mellitus is a metabolic disease detection and correct diagnosis of diabetes is important for
caused by high blood sugar, which can lead to serious the management and prevention of complications related to
health problems if not properly controlled. Early the disease.
prediction and timely intervention are crucial for
preventing and managing diabetes. This paper presents Machine learning has become a powerful tool for
a Diabetic Prediction System utilizing the Support predictive analytics and decision support in healthcare.
Vector Machine (SVM) algorithm, a powerful machine Support Vector Machine (SVM) is one of the machine
learning technique known for its effectiveness in learning algorithms that has been shown to be effective in
classification tasks. The proposed system lever- ages a task classification. In diabetes prediction, SVM can be used
dataset comprising relevant features such as age, body to analyze and classify patient data to help identify
mass index (BMI), family history, and blood pressure to individuals at risk for diabetes. Diabetes prediction using
train the SVM model. Data were preprocessed to the SVM algorithm aims to leverage the power of SVM to
control for missing values, normalize features, and create a powerful and accurate prediction model to identify
reduce bias. The SVM algorithm is employed for individuals at risk of diabetes. The system uses data
classification, as it excels in handling high-dimensional including important characteristics such as age, body mass
data and is capable of finding optimal hyperplanes to index (BMI), family history and diabetes level to train the
separate different classes. The system undergoes a SVM algorithm. Training models can be used to predict a
comprehensive evaluation using performance metrics person's risk of developing diabetes based on their input.
such as accuracy, sensitivity, specificity, and area under
the receiver operating characteristic curve. The results A. Motivation
demonstrate the efficacy of the SVM algorithm in The motivation for developing diabetes prediction
accurately predicting the likelihood of diabetes based on using the Support Vector Machine (SVM) algorithm stems
the input features. from the urgent need for efficient and effective methods to
control global health problems caused by diabetes. Diabetes
Keywords:- Support Vector Machine (SVM), Prediction has become a global epidemic in which people are
System, Machine Learning, Classification, Feature constantly affected. Early detection and management can
Selection. significantly reduce the risk of these complications, thereby
lessening the burden on healthcare systems. Predictive
I. INTRODUCTION systems can identify at risk individuals before symptoms
manifest, enabling proactive healthcare measures.
Diabetes is a chronic disease. If blood sugar is higher Providing healthcare professionals with a reliable
than normal, diabetes is diagnosed because there is a high predictive tool enhances their ability to make informed
and toxic insulin release. Diabetes causes many harms to decisions.
our body and causes the body's tissues, kidneys, eyes and
blood vessels to fail. Identifying this disease at an early B. Objectives
stage can help professionals worldwide prevent injuries. To develop and use prediction models based on the
We can divide diabetes into two main groups: type 1 support vector machine (SVM) algorithm to accurately
diabetes and type 2 diabetes. Common symptoms include predict diabetes risk. To collect relevant and
thirst and frequent urination. Because this type of diabetes comprehensive data related to diabetes risk factors,
must be treated, it cannot be eliminated with medication. including demo graphic information, medical history, and
Type 2 diabetes often occurs in the elderly and elderly and diagnostic test results. To identify and select the most
can lead to high blood pressure, obesity and other diseases. relevant features that significantly contribute to diabetes
Diabetes is the leading cause of death. What is needed is prediction. To demonstrate the practical utility of the
early detection and diagnosis of diabetes. The main issues Diabetic Prediction System in facilitating preventive
of classification are the diagnosis of diabetes and the measures.
interpretation of diabetes data.

Diabetes Mellitus is a metabolic disease caused by

high blood sugar due to insufficient insulin production or
inadequate insulin use. Diabetes is increasing worldwide
and is becoming a major public health problem. Early

IJISRT24MAY1274 www.ijisrt.com 2082

Volume 9, Issue 5, May – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAY1274

II. LITERATURE REVIEW into XGBoost to create features, then a prediction model is
created for risk classification, and finally high and low risk
A. Prognostication and Outcome Specific Risk Factor regression analysis is performed. Get a more accurate
Identification for Diabetes Care via Private-Shared model of type II diabetes after testing.
Multi-Task Learning
Diabetes is a chronic disease that affects D. DMNet: A Personalized Risk Assessment Framework
approximately 500 million people worldwide and is almost for Elderly
always associated with many complications, including Type 2 diabetes is the most common disease among
kidney failure, blindness, stroke, and heart disease. An adults. This disease is difficult to treat and causes ongoing
important step in improving diabetes treatment is to medical costs. Early and individual risk assessment for type
accurately estimate the risk of diabetes complications and 2 diabetes is necessary. To date, many methods have been
identify factors associated with the development of each proposed to estimate the risk of type 2 diabetes. But these
complication. In this article, we examine problem gambling methods have three main problems: 1) they do not take into
and a special phenomenon, diagnostic risk, from historical account the importance of personal and health information,
data. We adopt a multi-task learning (MTL) model that 2) they do not work long-term, and 3) they are not all
jointly models many problems, where each task relationships. is at risk of diabetes. Individual risk
corresponds to the risk model of a problem. The MTL assessment procedures for elderly patients with type 2
model not only improves prediction performance but also diabetes are needed to address these issues. But this is very
allows the identification of specific risk factors. difficult for two reasons: uneven label distribution and
Specifically, we decompose the coefficient matrix into pressure characteristics. In this paper, we propose a
shared elements and specific values of private elements, diabetes network framework (DMNet) for type 2
where each row (vector) corresponds to the coefficient of assessment of type 2 diabetes in adults. In particular, we
the risk model. propose consolidated long-term memory to extract long-
term information for different diabetes. Additionally, a
B. Machine learning tools to predict long-term risk of type tandem mechanism was used to capture the relationship
2 diabetes between diabetes.
The proportion of seniors who are willing and able to
contribute to society is constantly increasing. Therefore, E. Machine Learning-based Risk Prediction for type 2
early retirement or exit from the labor market, due to Diabetes (T2DM)
health-related issues, poses a significant problem. Today, Early diagnosis of people at highest risk of diabetes is
due to the advancement of technology and the increasing important to prevent the occurrence and development of the
amount of data coming from different cultures, research disease. Therefore, we plan to develop a predictive
and analysis of health problems are moving towards application for screening high-risk groups for type 2
automation. Within the scope of this study, a frame bag that diabetes (T2DM). Against this background, we designed
is worker-oriented, IoT-enabled, inconspicuous, capable of and conducted a survey-based cross-sectional study using
monitoring the user's health, health and work, and equipped traditional diabetes risk factors to examine the prevalence
with smart equipment is ready. Diabetes Mellitus is a and the relationship between occurrence and exposure. We
chronic disease that significantly affects quality of life and used chi-square tests and binary logistic regression to
mortality in developed and developing countries evaluate and analyze the most important factors of diabetes
worldwide. Therefore, its serious impact on people's lives risk for T2DM prediction. Synthetic minority
(personal, social, work, etc.) can be reduced if diagnosed oversampling. We used the same class data to examine the
early, but most research in this area does not offer a more best performance of the classification system to identify
personalized approach to modeling and prediction. In this patients at risk for diabetes with higher F1 scores.
introduction, we develop a system for estimating diabetes Hyperparameters of the best-performing models were
risk in which specific components of the knowledge further tuned using 10-fold cross-validation to obtain better
discovery process (KDD) are used, evaluated, and F1 scores.
calculated. Specifically, consider using different machine
learning (ML) models for data generation, feature selection, F. Diabetes Predicting mHealth Application Using
and classification. The ensemble Weighted Voting LRRFs Machine Learning.
ML model is proposed to improve the prediction of With the development of information technology,
diabetes, scoring an Area Under the ROC Curve (AUC) of mobile health (mHealth) technology can be used for patient
0.884. self-management, patient diagnosis, and determination of
the consequences of some diseases. Diabetes is a lifelong
C. Prediction of Type II Diabetes Risk Based on XGBoost disease that affects millions of people worldwide. Although
and 1DCNN there are some mobile applications that track calories,
This article uses machine learning techniques to health, medications, lifestyle, diabetes, blood pressure,
accurately measure blood glucose based on real physical personal weight and provide recommendations on diet and
examination data from a tertiary hospital. The raw data is exercise to prevent or control diabetes, there are no
preprocessed by a number of methods and then some published studies that clearly show the risk of developing
irrelevant features are removed by correlating the feature diabetes. Therefore, the goal of this article is to create a
values with the target value. Feature data is then entered machine learning-based smartphone medical application

IJISRT24MAY1274 www.ijisrt.com 2083

Volume 9, Issue 5, May – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAY1274

that can evaluate the likelihood of blood sugar or non- The updated dataset under consideration includes the
diabetic conditions being diabetes without the help of a following attributes: gender, age, heart disease,
doctor or medical examination. hypertension, smoking history, BMI, hemoglobin A1c
(HbA1c) level, glucose level, and outcome. The proposed
G. Web Application-based Diabetes Prediction using system takes into account patients who are younger than
Machine Learning. 21.
Diabetes is a serious disease that many people
struggle with. People with diabetes are at high risk for heart
disease, kidney disease, stroke, eye problems, tissue
damage and more. The current practice of hospitals is to
collect the necessary information about diabetes problems
by diagnosing the disease through different diagnostic tests
and provide appropriate treatment according to the
diagnostic pain, which requires more strength and skill.
However, this big problem can be solved using machine
learning. Machine learning algorithms K-nearest neighbor
(KNN) and random forest (RF) are used to predict diabetes
risk in planning studies. After preliminary data, features are
selected based on their relevance to disease prediction.
After feature selection and clustering, the prediction
accuracy using Random Forest (RF) on previous data is
75% better than K-Nearest Neighbors (KNN). The goal is
to predict diabetes risk using machine learning techniques
and create a web application to support this prediction.

H. Artificial Intelligence Enabled Web-Based Prediction of Fig 1: Proposed System Block Diagram
Diabetes using Machine Learning Approach
Medical care is health care that includes diagnosis, B. Algorithm
surgery, treatment, treatment and other activities related to
 Support Vector Machine
human health. With the advancement of technology,
Develop a system that can accurately classify leaf
medical care has been improved with smart medicine, e-
medicine and cell therapy applications. In recent years, images as healthy or diseased. Support vector machine
computer scientists have become interested in improving (SVM) is a supervised machine learning algorithm that can
human health, which requires extensive research on be used for classification or regression. The main purpose
emerging diseases in clinical decision-making. In this of SVM is to find the plane that best divides the points into
study, early diabetes risk information was trained with different groups. In two-dimensional space, a hyperplane is
supervised machine learning and classified with a line that divides data into two groups.
unsupervised machine learning. Classification of diabetes
based on best accuracy of supervised machine learning Support vectors are the data points that are closest to
the hyperplane and have the maximum margin. SVM aims
algorithm for novel diagnosis. Create a web app to predict
to maximize this margin as it generally leads to better
early diabetes risk by classifying results based on patient
questions without using machine learning for labs. generalization performance on unseen data.
Additionally, the results were analyzed with unsupervised
machine learning and grouped according to the likelihood SVM inherently supports binary classification. For
of predicting positive or negative blood sugar levels. multiclass problems, techniques such as one-vs-one or one-
Review evaluation of predictions from deep learning to vs-all are commonly used. Assemble a comprehensive
improve accuracy. dataset of leaf images, encompassing various crops and
disease types. Resize, normalize, and preprocess the dataset
for effective model training. Extract meaningful features
III. METHODOLOGY
from the leaf images to represent them in a format suitable
for SVM.
A. Proposed System
We use a single algorithm in the proposed system as
per shown in (figure 1), which lowers the time complexity. Design an SVM-based classification model for leaf
SVM (Support Vector Machine) is a machine learning disease detection. Train the SVM model using the training
technique used to predict diabetes. dataset, optimizing hyperparameters as needed. Utilize
metrics like accuracy, precision, recall, and F1-score to
We can take patient information into account evaluate the SVM model’s performance. Implement the
regardless of age and gender. The application process is an SVM model in a deployable format and integrate it into the
interactive application that requires users to enter data to user interface.
generate an estimate.

IJISRT24MAY1274 www.ijisrt.com 2084

Volume 9, Issue 5, May – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAY1274

 SVM Algorithm Working Steps: IV. IMPLEMENTATION AND RESULTS

 Step 1: Load the important libraries. We’re building a website that predicts the diabetes by
 Step 2: Import dataset and extract the X variables and using SVM machine learning algorithm. So, we gather
Y separately. datasets containing relevant features such as glucose levels,
 Step 3: Divide the dataset into train and test. BMI, age, etc., as well as data from medical questionnaires
 Step 4: Initializing the SVM classifier model. related to diabetes risk factors and clean the data to correct
 Step 5: Fitting the SVM classifier model. errors and handle missing values. Ensure data is in a format
 Step 6: Coming up with predictions. readable by computers.
 Step 7: Evaluating model’s performance.
We extract significant features from the data, such as
glucose levels, insulin levels, age, and family history of
The objective of the Support Vector Machine (SVM)
algorithm is to establish an optimal decision boundary, diabetes and applied preprocessing techniques like
known as a hyperplane, within an n-dimensional space to normalization or standardization to ensure features are on a
effectively separate different classes. This hyperplane similar scale. Then we utilize SVM algorithm for
classification to predict the likelihood of an individual
facilitates the categorization of new data points into the
having diabetes and train the SVM model using the selected
appropriate class in the future.
features from the dataset.
SVM selects point clouds/vectors that help create an
After that we evaluate the SVM model’s performance
overall plane. These conditions are called support vectors,
using metrics like accuracy, precision, recall, and F1-score.
so the algorithm is called a vector machine. Consider the
Employ techniques like cross-validation to assess the
below diagram (figure 2) in which there are two different
model’s generalization ability. Explore combining SVM
categories that are classified using a decision boundary or
with other algorithms or techniques, such as ensemble
hyperplane:
methods or feature selection algorithms, to enhance
prediction accuracy. Then we split the dataset into training,
validation, and test sets to train, tune hyperparameters, and
evaluate the model’s performance and develop a web-based
platform where individuals can input their relevant health
data, and the SVM model predicts their risk of developing
diabetes. Ensure the website is user-friendly and provides
clear explanations of the predictions.

We implement mechanisms to monitor model

performance and update the model periodically with new
data or improved algorithms and provide ongoing support
and maintenance for the website to ensure its functionality
and accuracy.

Fig 2: SVM Algorithm Working

Fig 3: Login Page

IJISRT24MAY1274 www.ijisrt.com 2085

Volume 9, Issue 5, May – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAY1274

Fig 4: New User Resgistration Page

Fig 5: Home Page

Fig 6: About Us Page

IJISRT24MAY1274 www.ijisrt.com 2086

Volume 9, Issue 5, May – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAY1274

Fig 7: Instructions to Control Diabetes Page

Fig 8: Prediction Page

IJISRT24MAY1274 www.ijisrt.com 2087

Volume 9, Issue 5, May – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAY1274

Fig 9: Result Page

Fig 10: Diet Plan Page

IJISRT24MAY1274 www.ijisrt.com 2088

Volume 9, Issue 5, May – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAY1274

Fig 11: Consult Page

V. CONCLUSION is due to the medical professionals and endocrinologists

who provided invaluable expertise and annotated datasets,
The Diabetic Prediction System utilizing the Support ensuring the reliability and clinical relevance of SVM-
Vector Machine (SVM) algorithm represents a significant based models. Furthermore, acknowledgment extends to
stride in the field of predictive healthcare, specifically the patients who consented to share their medical data,
aimed at early identification and proactive management of enabling the training and validation of these algorithms on
individuals at risk of developing diabetes. This system diverse patient populations. Moreover, support from
harnesses the power of advanced machine learning to healthcare institutions, funding agencies, and industry
provide accurate and interpretable predictions, contributing collaborators has been indispensable in driving the
to im- proved patient outcomes, cost- efficient healthcare, translation of SVM-based diabetes prediction from research
and a positive impact on public health. laboratories to clinical practice. Ultimately, the
acknowledgment underscores the collective endeavor of
The Diabetic Prediction System using SVM multidisciplinary teams committed to harnessing the power
Algorithm stands as a promising tool in the pursuit of of SVM for the benefit of diabetic care and patient
proactive and personalized healthcare. By combining outcomes.
sophisticated machine learning techniques with a
thoughtful integration into healthcare workflows, this REFERENCES
system has the potential to significantly improve patient
outcomes and contribute to the broader goals of public [1]. American Diabetes Association. Economic costs of
health. diabetes in the u.s. in 2020. Diabetes Care,
41(5):917–928, 2021. American Diabetes
ACKNOWLEDGMENT Association and others. Expert Panel Report Most
Popular Articles for Scientific Research. Diabetes
Acknowledgment for the development and care, 26(suppl 1): s5–s20, 2022.
implementation of Support Vector Machine (SVM) models [2]. B. Liu, Y. Li, S. Ghosh, Z. Sun, K. Ng, and J. Hu.
in diabetes prediction represents a recognition of the Risk assessment in diabetes care: Bayesian
collaborative efforts and advancements in both medical and multitasking and social theory. IEEE Transactions
technological domains. Firstly, acknowledgment extends to on Knowl- edge and Data Engineering, 32(7):1276–
the researchers and data scientists who pioneered the 1289, 2020.
application of SVM in diabetes diagnosis, pushing the [3]. B. Kalaiselvi, “Improving random forest distribution
boundaries of AI-driven healthcare analytics. Their tireless based on human relations effectiveness for
work in developing and fine-tuning these algorithms has technology prediction models.,” Measurement, vol.
paved the way for more accurate and efficient prediction of 162, Oct. 2020, Art. no. 107885.
diabetes onset and progression. Additionally, appreciation

IJISRT24MAY1274 www.ijisrt.com 2089

Volume 9, Issue 5, May – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAY1274

[4]. R. Muthukrishnan and R. Rohini, “LASSO: A

feature selection technique in predictive modeling
for machine learning,” in Proc. IEEE Int. Conf. Adv.
Comput. Appl. (ICACA), Oct. 2020, pp. 18–20.
[5]. N. Long and S. Dagogo-Jack. Comorbidities of
diabetes and high blood pressure: mechanisms and
technique to target organ safety. The Journal of
Clinical Hypertension, 13(4):244–251, 2021.
[6]. W. Engchuan, A. C. Dimopoulos, S. Tyrovolas, F.
F. Caballero, A. Sanchez-Niubo, H. Arndt, J. L.
Ayuso-Mateos, J. M. Haro, S. Chatterji, and D. B.
Panagiotakos. Med. Sci. Monitor Int. Med. J. Exp.
Clin. Res., vol. 25, p. 1994, Mar. 2021.
[7]. J. Yanase and E. Triantaphyllou, “A systematic
survey of laptop-aided prognosis in medicinal drug:
beyond and present tendencies,” Expert Syst. Appl.,
vol. 138, Dec. 2019, Art. no. 112821.
[8]. D. Goksuluk, S. Korkmaz, G. Zararsiz, and E.
Karaagaoglu, “easyROC: An interactive web-tool
for ROC curve analysis using R language
environment,” R J., vol. 8, pp. 213–230, Dec. 2021.
[9]. Z. He and W. Yu, “Stable function selection for
biomarker find out,” Comput. Biol. Chem., vol. 34,
no. 4, pp. 215–225, 2020.

IJISRT24MAY1274 www.ijisrt.com 2090