Diabetes Prediction using Machine Learning Algorithms

Anirudha Gondhali, Dhanashree Daphale, Omkar khedekar.

“
MARATHAWADA MITRA MANDALS COLLEGE OF ENGINEERING, PUNE”

Abstract

Diabetes Mellitus is among critical diseases and lots of people are suffering from this disease. Age, obesity, lack of
exercise, hereditary diabetes, living style, bad diet, high blood pressure, etc. can cause Diabetes Mellitus. People
having diabetes have high risk of diseases like heart disease, kidney disease, stroke, eye problem, nerve damage, etc.
Current practice in hospital is to collect required information for diabetes diagnosis through various tests and
appropriate treatment is provided based on diagnosis. Big Data Analytics plays an significant role in healthcare
industries. Healthcare industries have large volume databases. Using big data analytics one can study huge datasets
and find hidden information, hidden patterns to discover knowledge from the data and predict outcomes accordingly.
In existing method, the classification and prediction accuracy is not so high. In this paper, we have proposed a
diabetes prediction model for better classification of diabetes which includes few external factors responsible for
diabetes along with regular factors like Glucose, BMI, Age, Insulin, etc. Classification accuracy is boosted with new
dataset compared to existing dataset. Further with imposed a pipeline model for diabetes prediction intended towards
improving the accuracy of classification.

I.INTRODUCTION
Healthcare sectors have large volume databases. Such databases may contain structured, semi-structured or
unstructured data. Big data analytics is the process which analyses huge data sets and reveals hidden information,
hidden patterns to discover knowledge from the given data.
Considering the current scenario, in developing countries like India, Diabetic Mellitus (DM) has become a very
severe disease. Diabetic Mellitus (DM) is classified as Non-Communicable Disease (NCB) and many people are
suffering from it. Around 425 million people suffer from diabetes according to 2017 statistics. Approximately 2-5
million patients every year lose their lives due to diabetes. It is said that by 2045 this will rise to 629 million.[1]
Diabetes Mellitus (DM) is classified as-
Type-1 known as Insulin-Dependent Diabetes Mellitus (IDDM). Inability of human’s body to generate sufficient
insulin is the reason behind this type of DM and hence it is required to inject insulin to a patient. Type-2 also known
as Non-Insulin-Dependent Diabetes Mellitus (NIDDM). This type of Diabetes is seen when body cells are not able
to use insulin properly.Type-3 Gestational Diabetes, increase in blood sugar level in pregnant woman where
diabetes is not detected earlier results in this type of diabetes. DM has long term complications associated with it.
Also, there are high risks of various health problems for a diabetic person.
A technique called, Predictive Analysis, incorporates a variety of machine learning algorithms, data mining
techniques and statistical methods that uses current and past data to find knowledge and predict future events. By
applying predictive analysis on healthcare data, significant decisions can be taken and predictions can be made.
Predictive analytics can be done using machine learning and regression technique. Predictive analytics aims at
diagnosing the disease with best possible accuracy, enhancing patient care, optimizing resources along with
improving clinical outcomes.[1] Machine learning is considered to be one of the most important artificial
intelligence features supports development of computer systems having the ability to acquire knowledge from past
experiences with no need of programming for every case. Machine learning is considered to be a dire need of
today’s situation in order to eliminate human efforts by supporting automation with minimum flaws. Existing
method for diabetes detection is uses lab tests such as fasting blood glucose and oral glucose tolerance. However,
this method is time consuming. This paper focuses on building predictive model using machine learning algorithms
and data mining techniques for diabetes prediction. The paper is organized as follows-
Section II-gives literature review of the work done on diabetes prediction earlier and taxonomy of machine learning
algorithms. Section III-presents motivation behind working on this topic. Section IV gives diabetes prediction
proposed model is discussed. Section V gives results of experiment followed by Conclusion and References.
 Diabetes prediction using machine learning

II.LITERATURE REVIEW
The analysis of related work gives results on various healthcare datasets, where analysis and predictions were carried
out using various methods and techniques. Various prediction models have been developed and implemented by
various researchers using variants of data mining techniques, machine learning algorithms or also combination of
these techniques. Dr Saravana Kumar N M, Eswari, Sampath P and Lavanya S (2015) implemented a system using
Hadoop and Map Reduce technique for analysis of Diabetic data. This system predicts type of diabetes and also
risks associated with it. The system is Hadoop based and is economical for any healthcare organization.[4]
Aiswarya Iyer (2015) used classification technique to study hidden patterns in diabetes dataset. Naïve Bayes and
Decision Trees were used in this model. Comparison was made for performance of both algorithms and
effectiveness of both algorithms was shown as a result.[5] K. Rajesh and V. Sangeetha (2012) used classification
technique. They used C4.5 decision tree algorithm to find hidden patterns from the dataset for classifying efficiently.
[8] Humar Kahramanli and Novruz Allahverdi (2008) used Artificial neural network (ANN) in combination with
fuzzy logic to predict diabetes.[9] B.M. Patil, R.C. Joshi and Durga Toshniwal (2010) proposed Hybrid Prediction
Model which includes Simple K-means clustering algorithm, followed by application of classification algorithm to
the result obtained from clustering algorithm. In order to build classifiers C4.5 decision tree algorithm is used.[10]
Mani Butwall and Shraddha Kumar (2015) proposed a model using Random Forest Classifier to forecast diabetes
behaviour.[7] Nawaz Mohamudally1 and Dost Muhammad (2011) used C4.5 decision
tree algorithm, Neural Network, K-means clustering algorithm and Visualization to predict diabetes.[11]
Fig 1, represents taxonomy for Machine Learning Algorithms that can be used for diabetes prediction.
The task of choosing a machine learning algorithm includes feature matching of the data to be learned based on
existing approaches. Taxonomy of machine learning algorithms is discussed below-
Machine learning has numerous algorithms which are classified into three categories: Supervised learning,
Unsupervised learning, Semi-supervised learning.

Fig1. Taxonomy of Machine Learning Algorithms for Diabetes Prediction

A.The Supervised Learning/Predictive Models

Supervised learning algorithms are used to construct predictive models. A predictive model predicts missing value
using other values present in the dataset. Supervised learning algorithm has a set of input data and also a set of
output, and builds a model to make realistic predictions for the response to new dataset. Supervised learning
includes Decision Tree, Bayesian Method, Artificial Neural Network, Instance based learning, Ensemble Method.
These are booming techniques in Machine learning.[3]

B. Unsupervised Learning / Descriptive Models

Descriptive models are developed using unsupervised learning method. In this model we have known set of inputs
but output is unknown. Unsupervised learning is mostly used on transactional data. This method includes clustering
algorithms like k-Means clustering and k-Medians clustering.[3]
 Diabetes prediction using machine learning

C. Semi-supervised Learning
Semi Supervised learning method uses both labeled and unlabeled data on training dataset. Classification,
Regression techniques come under Semi Supervised Learning. Logistic Regression, Linear Regression are examples
of regression techniques.[3]

III. MOTIVATION
There has been drastic increase in rate of people suffering from diabetes since a decade. Current human lifestyle is
the main reason behind growth in diabetes. In current medical diagnosis method, there can be three different types of
errors-
1. The false-negative type in which a patient in reality is already a diabetic patient but test results tell that the
person is not having diabetes.
2. The false-positive type. In this type, patient in reality is not a diabetic patient but test reports say that he/she
is a diabetic patient.
3. The third type is unclassifiable type in which a system cannot diagnose a given case. This happens due to
insufficient knowledge extraction from past data, a given patient may get predicted in an unclassified type.
However, in reality, the patient must predict either to be in diabetic category or non-diabetic category. Such errors
in diagnosis may lead to unnecessary treatments or no treatments at all when required. In order to avoid or reduce
severity of such impact, there is a need to create a system using machine learning algorithm and data mining
techniques which will provide accurate results and reduce human efforts.

IV.PROPOSED METHOD

Fig 2. Diabetes Prediction Model

Fig 2., shows architecture diagram for diabetes prediction model. This model has five different modules. These
modules include-
i. Dataset Collection
ii. Data Pre-processing
iii. Clustering
iv. Build Model
v. Evaluation

Let’s have a look at each model briefly.

i. Dataset Collection
 Diabetes prediction using machine learning
This module includes data collection and understanding the data to study the patterns and trends which helps in
prediction and evaluating the results. Dataset description is given below- This Diabetes dataset contains 800 records
and 10 attributes.
Table 1. Dataset Information
Attributes Type
Number of Pregnancies N
Glucose Level N

Blood Pressure N
Skin Thickness(mm) N
Insulin N
BMI N
Age N
Job Type(Office-work/Fieldwork/Machine-work) No

Outcome C

ii. Data Pre-processing

This phase of model handles inconsistent data in order to get more accurate and precise results. This dataset contains
missing values. So we imputed missing values for few selected attributes like Glucose level, Blood Pressure, Skin
Thickness, BMI and Age because these attributes cannot have values zero. Then we scale the dataset to normalize all
values.

iii. Clustering
In this phase, we have implemented K-means clustering on the dataset to classify each patient into either a diabetic
or non-diabetic class. Before performing K-means clustering, highly correlated attributes were found which were,
Glucose and Age. K-means clustering was performed on these two attributes. After implementation of this clustering
we got class labels (0 or 1) for each of our record.

Algorithm:
• Choose the number of clusters(K) and obtain the data points
• Place the centroids c_1, c_2, ..... c_k randomly
• Steps 4 and 5 should be repeated until the end of a fixed number of iterations  For each data point x_i:
-find the nearest centroid(c_1, c_2 ,..c_k)
-assign the point to that cluster
• for each cluster j = 1..k new centroid = mean of all points assigned to that cluster
• End
iv.Model Building
This is most important phase which includes model building for prediction of diabetes. In this we have implemented
various machine learning algorithms for diabetes prediction. These algorithms include Support Vector Classifier,
Random Forest Classifier, Decision Tree Classifier, Extra Tree Classifier, Ada Boost algorithm, Perceptron, Linear
Discriminant Analysis algorithm, Logistic Regression, K-Nearest Neighbour, Gaussian Naïve Bayes, Bagging
algorithm, Gradient Boost Classifier.

Algorithm 1: Diabetes Prediction using various machine learning algorithms

Generate training set and test set randomly. Specify algorithms that are used in
model mn=[ KNN( ), DTC( ), GaussianNB( ),
LDA(),SVC(),LinearSVC(),AdaBoost(), RandomForestClassifier(), Perceptron(),
ExtraTreeClassifier(), Bagging(),
 Diabetes prediction using machine learning
LogisticRegression(),
GradientBoostClassifier()]
for(i=0; i<13; i++) do
Model= mn[i]; Model.fit(); model.predict();
print(Accuracy(i),confusion_matrix, classification_report);
End

v.Evaluation
This is the final step of prediction model. Here, we evaluate the prediction results using various evaluation metrics
like classification accuracy, confusion matrix and f1-score.
Classification Accuracy- It is the ratio of number of correct predictions to the total number of input samples. It is
given as-

Confusion Matrix- It gives us gives us a matrix as output and describes the complete performance of the model.

Where, TP: True Positive

FP: False Positive
FN: False Negative
TN: True Negative

Accuracy for the matrix can be calculated by taking average of the values lying across the main diagonal. It is given
as-
Where, N:Total number of samples

F1 score-It is used to measure a test’s accuracy. F1 Score is the Harmonic Mean between precision and recall. The
range for F1 Score is [0, 1]. It tells you how precise your classifier is as well as how robust it is. Mathematically, it
is given as-

F1 Score tries to find the balance between precision and recall.

Precision: It is the number of correct positive results divided by the number of positive results predicted by the
classifier. It is expressed as-

Recall: It is the number of correct positive results divided by the number of all relevant samples. In mathematical
form it is given as-
 Diabetes prediction using machine learning

V. RESULTS
After applying various Machine Learning Algorithms on dataset we got accuracies as mentioned below. Logistic
Regression gives highest accuracy of 96%.

Table 2. Accuracy Table

Algorithms Accuracy
Decision Tree 86%
Gaussian NB 93%
LDA 94%
SVC 60%
Random Forest 91%
Extra Trees 91%
AdaBoost 93%
Perceptron 76%
Logistic Regression 96%
Gradient Boost Classifier 93%
Bagging 90%
KNN 90%

Confusion Matrix for Logistic Regression is given below-

Table 3. Confusion Matrix for Logistic Regression

Diabetic Non-
Diabetic
Diabetic 93 5
Non- 4 138
Diabetic

The different performance measures that are being compared are Accuracy, F1-Score, Precision and Recall. The
Confusion matrix for the algorithm with highest accuracy is mentioned in Table 3.
 Diabetes prediction using machine learning
Visualization of these accuracies helps us to understand variations among them clearly.

Fig 3.Comparison of various machine learning algorithms based on accuracies

Table 4. Comparison between accuracies of PIMA Diabetes Dataset and Diabetes Dataset used in this paper

Algorithms Accuracy with PIMA Accuracy with Diabetes

Dataset Dataset used in this
paper
Logistic Regression 76% 96%
Gradient Boost Classifier 77% 93%
LDA 77% 94%
AdaBoost Classifier 77% 93%
Extra Trees Classifier 76% 91%
Gaussian NB 67% 93%
Bagging 75% 90%
Random Forest 72% 91%
Decision Tree 74% 86%
Perceptron 67% 76%
SVC 68% 60%
KNN 72% 90%

CONCLUSION
In this study, various machine learning algorithms are applied on the dataset and the classification has been done
using various algorithms of which Logistic Regression gives highest accuracy of 96%. Application of pipeline gave
AdaBoost classifier as best model with accuracy of 98.8%. We have seen comparison of machine learning algorithm
accuracies with two different datasets. It is clear that the model improves accuracy and precision of diabetes
prediction with this dataset compared to existing dataset. Further this work can be extended to find how likely
nondiabetic people can have diabetes in next few years.

REFERENCES
[1] Gauri D. Kalyankar, Shivananda R. Poojara and Nagaraj V. Dharwadkar,” Predictive Analysis of Diabetic Patient Data Using Machine
Learning and Hadoop”, International Conference On I-SMAC,978-1-5090-3243-3,2017.
[2] Ayush Anand and Divya Shakti,” Prediction of Diabetes Based on Personal Lifestyle Indicators”, 1st International Conference on Next
Generation Computing Technologies, 978-1-4673-6809-4, September 2015.
[3] B. Nithya and Dr. V. Ilango,” Predictive Analytics in Health Care Using Machine Learning Tools and Techniques”, International
Conference on Intelligent Computing and Control Systems, 978-1-5386-2745-7,2017.
[4] Dr Saravana kumar N M, Eswari T, Sampath P and Lavanya S,” Predictive Methodology for Diabetic Data Analysis in Big Data”, 2nd
International Symposium on Big Data and Cloud Computing,2015.
[5] Aiswarya Iyer, S. Jeyalatha and Ronak Sumbaly,” Diagnosis of Diabetes Using Classification Mining Techniques”, International
Journal of Data Mining & Knowledge Management Process (IJDKP) Vol.5, No.1, January 2015.
 Diabetes prediction using machine learning
[6] P. Suresh Kumar and S. Pranavi “Performance Analysis of Machine Learning Algorithms on Diabetes Dataset using Big Data
Analytics”, International Conference on Infocom Technologies and Unmanned Systems, 978-1-5386-0514-1, Dec. 18-20, 2017.
[7] Mani Butwall and Shraddha Kumar,” A Data Mining Approach for the Diagnosis of Diabetes Mellitus using Random Forest
Classifier”, International Journal of Computer Applications, Volume 120 - Number 8,2015.
[8] K. Rajesh and V. Sangeetha, “Application of Data Mining Methods and Techniques for Diabetes Diagnosis”, International Journal of
Engineering and Innovative Technology (IJEIT) Volume 2, Issue 3, September 2012.
[9]Humar Kahramanli and Novruz Allahverdi,”Design of a Hybrid System for the Diabetes and Heart Disease”, Expert Systems with
Applications: An International Journal, Volume 35 Issue 1-2, July, 2008.
[10] B.M. Patil, R.C. Joshi and Durga Toshniwal,”Association Rule for Classification of Type-2 Diabetic Patients”, ICMLC '10
Proceedings of the 2010 Second International Conference on Machine Learning and Computing, February 09 - 11, 2010.
[11] Dost Muhammad Khan1, Nawaz Mohamudally2, “An Integration of K-means and Decision Tree (ID3) towards a more Efficient Data
Mining Algorithm ”, Journal Of Computing, Volume 3, Issue 12, December 2011.