Diabetes Prediction Using Machine Learning KNN - Algorithm Technique
Diabetes Prediction Using Machine Learning KNN - Algorithm Technique
Volume 7, Issue 5, May – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Diabetes Prediction Using Machine Learning
KNN -Algorithm Technique
Dr. B. Premamayudu (Professor)1 , K. Muralikrishna2 , K. Pramodh 3 ,
1
Professor, Department of Information Technology, Vignan’s Foundation for Science,
Technology & Research , Guntur AP-522213, India
2,3
UG student, Department of Infotmation Technology Vignan’s Foundation for Science,
Technology & Research, Guntur AP-522213, India
Abstract:- Diabetes is a chronic disease caused due to high urge sugar from the blood into the cells. that results in
amount of glucose present in the human body. If this increase in blood glucose levels. Glucose, the shape of sugar
diabetes is ignored, this may lead to severe health found in your blood, is one amongst your main energy
problems such as kidney failure, heart attacks, blood sources. There area unit three main kinds of polygenic
pressure, eye damage, weight loss, frequent urination, etc. disorder they're
Basically, human body contains Insulin which is produced 1.Type one polygenic disorder.
by pancreas. This insulin helps to enter glucose in to blood 2.Type a pair of polygenic disorder.
cells in order to generate energy to the body. There are 3.Gestational polygenic disorder.
types in diabetes Type1 and Type 2 other form is
gestational diabetes which is caused during pregnancy. Type one diabetes:
This can be controlled in the earlier stages of the attack. It is believed to be reaction condition. this suggests your
According to International Diabetes Federation (IDF) 382 system erroneously attacks and destroys the beta cells in your
million people are suffering with diabetes and by next duct gland that produces hypoglycemic agent. The harm is
20years the count will be doubled as 592 million. To permanent. we tend to cannot realize the prompts of sickness
accomplish this goal, in this project we can do early simply. There could also be each genetic and environmental
prediction of diabetes in humans or patients for good reasons and modus vivendi factors thought to play a job.
accuracy through applying various machine learning
techniques such as Random Forest (RF), K-nearest Type a pair of diabetes:
neighbors (KNN), Decision Trees (DT), etc. However, in This type starts as hypoglycemic agent resistance. Our
this project we are predicting diabetes using KNN body cannot respond for systematic hypoglycemic agent.
classifier model. As we see now a days machine learning That regulates duct gland to supply additional hypoglycemic
is an emerging technology and boon to many problem agent because it isn't sensible for health. hypoglycemic agent
solutions. production decreases and results in high blood glucose levels.
The reason behind this kind sickness is genetic science, lack
I. INTRODUCTION of exercise, being overweight.
Machine Learning Gestational diabetes:
Machine learning (ML) could be a variety of AI (AI) This is thanks to hypoglycaemic agent obstruction
that enables code applications to become additional correct at hormones created by throughout maternity. this kind of
predicting outcomes while not being expressly programmed sickness happens solely throughout maternity solely.
to try and do, therefore. Machine learning algorithms use
historical knowledge as input to predict new output values. Symptoms:
Blood pressure downside repeated elimination
Types of Learning: Dry and fidgety skin
1. Supervised learning. Visionary issues
2. Unsupervised learning. Slow recovery of health conditions
3. Reinforcement learning
II. LITERATURE REVIEW
In this project, we have a tendency to square measure
victimization supervised learning classifier technique. i.e., 1. KM. Jyothirani aims to apply 5 machine learning
KNN algorithmic rule to search out the accuracy of predicting classification algorithms to predict diabetes and compare
the new outcomes. each to find which algorithm gives accurate target
outcomes. In her research PIMA datasets were used and
In this project we tend to use some datasets to predict the study concluded that Decision trees gave 98%
the attack of polygenic disorder to the folks. Diabetes could accuracy score.
be a fast-growing sickness in folks even in kids too. It's a 2. Avantika Nahar had applied the KNN algorithm for
gaggle of sickness during which blood doesn’t turn out classification and prediction of diabetes using trained data
enough quantity of hypoglycemic agent, doesn’t properly use and predicts the time of getting diabetes also. This project
the hypoglycemic agent that's created. The body is unable to
IJISRT22MAY682 www.ijisrt.com 941
Volume 7, Issue 5, May – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
result is based on YES or NO. if the result is NO then time Table 1: Dataset Description
prediction module is used. Else we use just prediction of S. No Attributes
diabetes and accuracy of the KNN algorithm. 1 Pregnancy
3. Umatejaswi and P. Suresh Kumar had talked about 2 Glucose
algorithms such as Support Vector Machine, NaiveBias,
Decision Trees in order to find diseases through data 3 Blood pressure
mining technique. 4 Skin thickness
5 Insulin
III. METHODOLOGY 6 BMI (body mass index)
7 Diabetes pedigree function
In this section, we are learning KNN classifier model
8 age
used in machine learning to predict diabetes. We shall also
explain our proposed methodology to improve the accuracy
The 9th attribute is class variable of each data points.
of finding the targeted outcomes.
This class variable shows the outcomes 0 & 1 for diabetes
which indicates non-diabetic & diabetic.
A. Dataset Description: -
This data is collected from UCI repository which is
named as PIMA Indian diabetes dataset. The dataset has
many attributes of 768 patients.
Dataset head part
Above values are the statistical values of the dataset Distribution of diabetic patient
which we have used. This model is made to predict how many numbers of
patients are having diabetes. In this below outcome we can
Here, from this correlation matrix we came to know that see 0’s label contains 500 classes and 1’s label contains 268
pregnancies and glucose columns are very important to classes.
predict the output. These two columns played key role.
IJISRT22MAY682 www.ijisrt.com 942
Volume 7, Issue 5, May – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
B. Data pre-processing: We have many classification techniques such as support
This is the most crucial process. Mostly healthcare vector machine (SVM), random forest, decision trees, KNN
related data may contain many missing values and many algorithm etc. However here in this project we are using K
mistakes which might cause for low effective of data.so to nearest neighbour’s classifier technique only.
improve the quality and effectiveness data processing should
be done. This process is more essential to get good accuracy. KNN Classifier
There are mainly two steps in this data pre-processing they It is one in every of the best machine learning algorithms
are supported supervised learning techniques.
1.Missing values removal.
It could be a non-parametric rule, which implies it doesn't
2.Splitting of data into training and testing sets.
build any assumption on underlying knowledge.
C. Applying classifier technique: It is additionally known as as lazy learner rule as a result
After training and testing datasets are separated and of it doesn't learn from the coaching set quickly.
without null values in the dataset, we can now apply the
machine learning classifier technique to the dataset.
IJISRT22MAY682 www.ijisrt.com 943
Volume 7, Issue 5, May – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
At the coaching part simply stores the knowledge set and the diabetes and save the lives of humans. if we would apply
once it gets new data, then it classifies that knowledge into this pattern in finding diabetics in patients it would be really
class that's abundant the same as the new knowledge. helpful for all the humans and hospital management as well.
We can find the results fast.
Working of KNN:
• Select the amount K of the neighbours. Output accuracy
• Calculate the euclidian distance of K variety of neighbors.
• Take the K neighbors as per the calculated euclidian
distance.
• Among these K neighbors, count the amount of the info
points in every class.
• Assign the new knowledge points there to class that the
amount of neighbors is most.
• our model is prepared to use.
REFERENCES
[1]. Mitushi soni, Dr. Sunitha Varma “Diabetes Prediction
victimization Machine Learning Techniques”
IJERTV9ISO90496.
[2]. Avantika Nahar, Dr. Ajay Lala, Saurabh Sharma,
“Diabetes Prediction victimization Machine Learning”
ISSN 2347-6435.
[3]. k. Jyothi aristocrat “Diabetes Prediction victimization
Machine Learning” IJSRCSEIT206463, ISSN: 2456-
3307.
[4]. Deeraj Shetty, Kishor Rit, Sohail Shaikh, Nikita Patil,
"Diabetes Prediction victimization data processing
"(ICIIECS), 2017
[5]. Vijayakumar, Lavanya, I. Nirmala, Sofia Carolingian,
"Random Forest algorithmic program for the Prediction
of polygenic disorder “, 2019.
[6]. Debadri Dutta, Debpriyo Paul, Parthajeet Ghosh,
"Analyzing Feature Importances for polygenic disorder
Prediction victimization Machine Learning". IEEE, pp
942-928, 2018.
[7]. A.K., Dewangan, and P., Agrawal, Classification of
diabetes victimization Machine Learning Techniques,
International Journal of Engineering and Applied
IV. EXPERIMENTAL RESULTS Sciences, vol. 2, 2015.
[8]. Nahla, Andrew – “Intelligible support vector machines
In this task different stages were performed. This for diagnosis of diabetes mellitus.” “Information
approach used K- Nearest Neighbors (KNN) classifier technology in biomedicine” IEEE transactions.
technique. Using this machine learning technique, we find the (July,2010),1114-20.
accuracy of predicting diabetes using KNN algorithm. And
we have got the accuracy score of 79% which is better to
apply for prediction. Overall, study states that we can use this
KNN algorithm for achieving high performance accuracy.
There are many variants in KNN algorithm, all those variants
may give different accuracy scores compared to the accuracy
which we got now.
V. CONCLUSION
The main target of this project was to find whether KNN
classifier algorithm is suitable for prediction or not. This we
can see by checking the performance analysis, which we had
get 79%. To find this accuracy we use the library called sci-
kit learn in python. This accuracy is good to apply for
prediction. The experimental results can be helpful in
healthcare to predict and make early decision-making to cure
