Introduction

The project, Diabetes Prediction Analysis Using Machine Learning in R, utilized the Pima Indians Diabetes dataset from Kaggle. This dataset consists of 765 rows and 9 features: pregnancies, glucose levels, blood pressure, skin thickness, insulin levels, BMI, diabetes pedigree function, age, and the outcome variable indicating diabetes presence. The goal was to evaluate and compare the performance of various machine learning models for diabetes prediction and identify key predictive features.

Key Achievements

Exploratory Data Analysis (EDA): Analyzed feature distributions, correlations, and identified key variables influencing diabetes outcomes. Applied normalization to numerical features and detected outliers for data quality improvement.
Model Implementation and Evaluation:

Logistic Regression: Achieved an accuracy of 79.13% and an AUC of 0.8436, indicating strong performance and discrimination capability.

Support Vector Machine (SVM): Delivered the highest accuracy of 80%, demonstrating its suitability for binary classification.

Random Forest: Provided an accuracy of 74.78%, along with feature importance insights for interpretability.

XGBoost: Achieved an accuracy of 73.04% and an AUC of 0.8124, highlighting its potential with optimized hyperparameters.

Performance Visualization: ROC curves were plotted for each model to compare their ability to distinguish between positive and negative diabetes cases, with Logistic Regression and SVM demonstrating superior results.

Conclusion

This analysis successfully demonstrates the application of machine learning techniques to predict diabetes outcomes using the Pima dataset. Logistic Regression stood out with a balance of high accuracy (79.13%) and AUC (0.8436), while SVM achieved the highest accuracy (80%). Random Forest and XGBoost offered complementary insights through feature importance analysis and robust ensemble learning techniques. The project underscores the importance of data-driven modeling in healthcare analytics and the predictive significance of features such as glucose levels and BMI in diagnosing diabetes.

options(repos = c(CRAN = "https://cloud.r-project.org"))

Install Required Packages

install.packages(c("dplyr", "ggplot2", "caret", "e1071", "randomForest", "pROC", "xgboost", "ggcorrplot"))

## Installing packages into 'C:/Users/OMPRAKASH/AppData/Local/R/win-library/4.4'
## (as 'lib' is unspecified)

## package 'dplyr' successfully unpacked and MD5 sums checked

## Warning: cannot remove prior installation of package 'dplyr'

## Warning in file.copy(savedcopy, lib, recursive = TRUE): problem copying
## C:\Users\OMPRAKASH\AppData\Local\R\win-library\4.4\00LOCK\dplyr\libs\x64\dplyr.dll
## to C:\Users\OMPRAKASH\AppData\Local\R\win-library\4.4\dplyr\libs\x64\dplyr.dll:
## Permission denied

## Warning: restored 'dplyr'

## package 'ggplot2' successfully unpacked and MD5 sums checked
## package 'caret' successfully unpacked and MD5 sums checked
## package 'e1071' successfully unpacked and MD5 sums checked
## package 'randomForest' successfully unpacked and MD5 sums checked
## package 'pROC' successfully unpacked and MD5 sums checked
## package 'xgboost' successfully unpacked and MD5 sums checked
## package 'ggcorrplot' successfully unpacked and MD5 sums checked
## 
## The downloaded binary packages are in
##  C:\Users\OMPRAKASH\AppData\Local\Temp\RtmpuSVJVU\downloaded_packages

Load the installed packages

library(dplyr)

## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union

library(ggplot2)
library(caret)

## Loading required package: lattice

library(e1071)
library(randomForest)

## randomForest 4.7-1.2

## Type rfNews() to see new features/changes/bug fixes.

## 
## Attaching package: 'randomForest'

## The following object is masked from 'package:ggplot2':
## 
##     margin

## The following object is masked from 'package:dplyr':
## 
##     combine

library(pROC)

## Type 'citation("pROC")' for a citation.

## 
## Attaching package: 'pROC'

## The following objects are masked from 'package:stats':
## 
##     cov, smooth, var

library(xgboost)

## 
## Attaching package: 'xgboost'

## The following object is masked from 'package:dplyr':
## 
##     slice

Load the dataset

data <- read.csv("diabetes.csv")

Inspect the dataset

head(data)

##   Pregnancies Glucose BloodPressure SkinThickness Insulin  BMI
## 1           6     148            72            35       0 33.6
## 2           1      85            66            29       0 26.6
## 3           8     183            64             0       0 23.3
## 4           1      89            66            23      94 28.1
## 5           0     137            40            35     168 43.1
## 6           5     116            74             0       0 25.6
##   DiabetesPedigreeFunction Age Outcome
## 1                    0.627  50       1
## 2                    0.351  31       0
## 3                    0.672  32       1
## 4                    0.167  21       0
## 5                    2.288  33       1
## 6                    0.201  30       0

summary(data)

##   Pregnancies        Glucose      BloodPressure    SkinThickness  
##  Min.   : 0.000   Min.   :  0.0   Min.   :  0.00   Min.   : 0.00  
##  1st Qu.: 1.000   1st Qu.: 99.0   1st Qu.: 62.00   1st Qu.: 0.00  
##  Median : 3.000   Median :117.0   Median : 72.00   Median :23.00  
##  Mean   : 3.845   Mean   :120.9   Mean   : 69.11   Mean   :20.54  
##  3rd Qu.: 6.000   3rd Qu.:140.2   3rd Qu.: 80.00   3rd Qu.:32.00  
##  Max.   :17.000   Max.   :199.0   Max.   :122.00   Max.   :99.00  
##     Insulin           BMI        DiabetesPedigreeFunction      Age       
##  Min.   :  0.0   Min.   : 0.00   Min.   :0.0780           Min.   :21.00  
##  1st Qu.:  0.0   1st Qu.:27.30   1st Qu.:0.2437           1st Qu.:24.00  
##  Median : 30.5   Median :32.00   Median :0.3725           Median :29.00  
##  Mean   : 79.8   Mean   :31.99   Mean   :0.4719           Mean   :33.24  
##  3rd Qu.:127.2   3rd Qu.:36.60   3rd Qu.:0.6262           3rd Qu.:41.00  
##  Max.   :846.0   Max.   :67.10   Max.   :2.4200           Max.   :81.00  
##     Outcome     
##  Min.   :0.000  
##  1st Qu.:0.000  
##  Median :0.000  
##  Mean   :0.349  
##  3rd Qu.:1.000  
##  Max.   :1.000

str(data)

## 'data.frame':    768 obs. of  9 variables:
##  $ Pregnancies             : int  6 1 8 1 0 5 3 10 2 8 ...
##  $ Glucose                 : int  148 85 183 89 137 116 78 115 197 125 ...
##  $ BloodPressure           : int  72 66 64 66 40 74 50 0 70 96 ...
##  $ SkinThickness           : int  35 29 0 23 35 0 32 0 45 0 ...
##  $ Insulin                 : int  0 0 0 94 168 0 88 0 543 0 ...
##  $ BMI                     : num  33.6 26.6 23.3 28.1 43.1 25.6 31 35.3 30.5 0 ...
##  $ DiabetesPedigreeFunction: num  0.627 0.351 0.672 0.167 2.288 ...
##  $ Age                     : int  50 31 32 21 33 30 26 29 53 54 ...
##  $ Outcome                 : int  1 0 1 0 1 0 1 0 1 1 ...

Check for missing values

sum(is.na(data))

## [1] 0

Normalize numerical columns (if needed)

data$Glucose <- scale(data$Glucose)
data$BMI <- scale(data$BMI)

Check for outliers using boxplots

boxplot(data$Age, main = "Age Outliers", horizontal=TRUE)

Correlation Matrix

library(ggcorrplot)
cor_matrix <- cor(data[,-ncol(data)]) # Exclude the target column
ggcorrplot(cor_matrix, method = "circle")

Glucose levels by Outcome

ggplot(data, aes(x = Outcome, y = Glucose, fill = as.factor(Outcome))) +
  geom_boxplot() +
  labs(title = "Glucose Levels by Outcome", x = "Outcome", y = "Glucose") +
  theme_minimal()

Install and load the caret package

if (!require(caret)) install.packages("caret")
library(caret)

# Split the data (70% train, 30% test)

set.seed(123)
index <- createDataPartition(data$Outcome, p = 0.7, list = FALSE)
train <- data[index, ]
test <- data[-index, ]

Logistic Regression Model

log_model <- glm(Outcome ~ ., data = train, family = "binomial")
summary(log_model)

## 
## Call:
## glm(formula = Outcome ~ ., family = "binomial", data = train)
## 
## Coefficients:
##                           Estimate Std. Error z value Pr(>|z|)    
## (Intercept)              -1.255871   0.511144  -2.457  0.01401 *  
## Pregnancies               0.103402   0.037988   2.722  0.00649 ** 
## Glucose                   1.144409   0.145945   7.841 4.46e-15 ***
## BloodPressure            -0.012665   0.006058  -2.091  0.03657 *  
## SkinThickness             0.003591   0.008093   0.444  0.65721    
## Insulin                  -0.001726   0.001060  -1.629  0.10335    
## BMI                       0.700192   0.141634   4.944 7.67e-07 ***
## DiabetesPedigreeFunction  0.699484   0.334854   2.089  0.03671 *  
## Age                       0.017113   0.011068   1.546  0.12206    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for binomial family taken to be 1)
## 
##     Null deviance: 695.03  on 537  degrees of freedom
## Residual deviance: 510.06  on 529  degrees of freedom
## AIC: 528.06
## 
## Number of Fisher Scoring iterations: 5

Predict and Evaluate

log_pred <- predict(log_model, test, type = "response")
log_class <- ifelse(log_pred > 0.5, 1, 0)

Confusion Matrix

conf_matrix <- table(test$Outcome, log_class)
conf_matrix

##    log_class
##       0   1
##   0 137  12
##   1  36  45

Accuracy

accuracy <- sum(diag(conf_matrix)) / sum(conf_matrix)
cat("Logistic Regression Accuracy:", accuracy)

## Logistic Regression Accuracy: 0.7913043

AUC

if (!require(pROC)) install.packages("pROC")
library(pROC)

roc_curve <- roc(test$Outcome, log_pred)

## Setting levels: control = 0, case = 1

## Setting direction: controls < cases

auc(roc_curve)

## Area under the curve: 0.8436

train$Outcome <- as.factor(train$Outcome)
test$Outcome <- as.factor(test$Outcome)

Train Random Forest model

if (!require(randomForest)) install.packages("randomForest")
library(randomForest)

rf_model <- randomForest(Outcome ~ ., data = train, ntree = 100, importance = TRUE)

Predict

rf_pred <- predict(rf_model, test)

Confusion Matrix

conf_matrix_rf <- table(test$Outcome, rf_pred)
print(conf_matrix_rf)

##    rf_pred
##       0   1
##   0 128  21
##   1  37  44

Accuracy

accuracy_rf <- sum(diag(conf_matrix_rf)) / sum(conf_matrix_rf)
cat("Random Forest Accuracy:", accuracy_rf)

## Random Forest Accuracy: 0.7478261

Feature Importance

importance(rf_model)

##                                   0           1 MeanDecreaseAccuracy
## Pregnancies               2.2039268 -0.04281905            1.7609905
## Glucose                  13.8313610 12.17217549           18.7840554
## BloodPressure             1.3734283  0.71549313            1.6741987
## SkinThickness             0.5786392  0.34728745            0.7102666
## Insulin                   2.5160861 -0.43223298            1.6246152
## BMI                       5.7419414  7.80288341            9.0527052
## DiabetesPedigreeFunction  2.5389156  1.41219476            2.7718543
## Age                       5.2459279  2.59371531            6.2453208
##                          MeanDecreaseGini
## Pregnancies                      19.39819
## Glucose                          60.60511
## BloodPressure                    23.00512
## SkinThickness                    17.95806
## Insulin                          18.16868
## BMI                              43.28194
## DiabetesPedigreeFunction         30.05861
## Age                              30.05910

varImpPlot(rf_model)

SVM Model

if (!require(e1071)) install.packages("e1071")
library(e1071)

svm_model <- svm(Outcome ~ ., data = train, kernel = "linear")
svm_pred <- predict(svm_model, test)

Confusion Matrix

conf_matrix_svm <- table(test$Outcome, svm_pred)
conf_matrix_svm

##    svm_pred
##       0   1
##   0 138  11
##   1  35  46

Accuracy

accuracy_svm <- sum(diag(conf_matrix_svm)) / sum(conf_matrix_svm)
cat("SVM Accuracy:",accuracy_svm)

## SVM Accuracy: 0.8

Prepare data for XGBoost

train_x <- as.matrix(train[, -ncol(train)])
train_y <- as.numeric(train$Outcome) - 1  # Convert to numeric and ensure labels are 0 or 1
test_x <- as.matrix(test[, -ncol(test)])
test_y <- as.numeric(test$Outcome) - 1    # Convert to numeric and ensure labels are 0 or 1

XGBoost Model

if (!require(xgboost)) install.packages("xgboost")
library(xgboost)

xgb_model <- xgboost(data = train_x, label = train_y, max_depth = 6, eta = 0.3, nrounds = 100, objective = "binary:logistic")

## [1]  train-logloss:0.558520 
## [2]  train-logloss:0.474518 
## [3]  train-logloss:0.413514 
## [4]  train-logloss:0.365560 
## [5]  train-logloss:0.330533 
## [6]  train-logloss:0.304606 
## [7]  train-logloss:0.281495 
## [8]  train-logloss:0.263482 
## [9]  train-logloss:0.240190 
## [10] train-logloss:0.225045 
## [11] train-logloss:0.214879 
## [12] train-logloss:0.202567 
## [13] train-logloss:0.191022 
## [14] train-logloss:0.183963 
## [15] train-logloss:0.176286 
## [16] train-logloss:0.168178 
## [17] train-logloss:0.163284 
## [18] train-logloss:0.153815 
## [19] train-logloss:0.147593 
## [20] train-logloss:0.143519 
## [21] train-logloss:0.140936 
## [22] train-logloss:0.136445 
## [23] train-logloss:0.131629 
## [24] train-logloss:0.125565 
## [25] train-logloss:0.122269 
## [26] train-logloss:0.119444 
## [27] train-logloss:0.113924 
## [28] train-logloss:0.111292 
## [29] train-logloss:0.107750 
## [30] train-logloss:0.104868 
## [31] train-logloss:0.100684 
## [32] train-logloss:0.097544 
## [33] train-logloss:0.095504 
## [34] train-logloss:0.092745 
## [35] train-logloss:0.088544 
## [36] train-logloss:0.085663 
## [37] train-logloss:0.083809 
## [38] train-logloss:0.080562 
## [39] train-logloss:0.077822 
## [40] train-logloss:0.075478 
## [41] train-logloss:0.073023 
## [42] train-logloss:0.070330 
## [43] train-logloss:0.068061 
## [44] train-logloss:0.067163 
## [45] train-logloss:0.065625 
## [46] train-logloss:0.064068 
## [47] train-logloss:0.062812 
## [48] train-logloss:0.061643 
## [49] train-logloss:0.060666 
## [50] train-logloss:0.059335 
## [51] train-logloss:0.058149 
## [52] train-logloss:0.057072 
## [53] train-logloss:0.055474 
## [54] train-logloss:0.054254 
## [55] train-logloss:0.052966 
## [56] train-logloss:0.052247 
## [57] train-logloss:0.051810 
## [58] train-logloss:0.050690 
## [59] train-logloss:0.049826 
## [60] train-logloss:0.049205 
## [61] train-logloss:0.048163 
## [62] train-logloss:0.047480 
## [63] train-logloss:0.046795 
## [64] train-logloss:0.045722 
## [65] train-logloss:0.044930 
## [66] train-logloss:0.044132 
## [67] train-logloss:0.043807 
## [68] train-logloss:0.043040 
## [69] train-logloss:0.042120 
## [70] train-logloss:0.041507 
## [71] train-logloss:0.040606 
## [72] train-logloss:0.040271 
## [73] train-logloss:0.039667 
## [74] train-logloss:0.039080 
## [75] train-logloss:0.038386 
## [76] train-logloss:0.038004 
## [77] train-logloss:0.037651 
## [78] train-logloss:0.037095 
## [79] train-logloss:0.036700 
## [80] train-logloss:0.036346 
## [81] train-logloss:0.035912 
## [82] train-logloss:0.035452 
## [83] train-logloss:0.034965 
## [84] train-logloss:0.034525 
## [85] train-logloss:0.034048 
## [86] train-logloss:0.033538 
## [87] train-logloss:0.033192 
## [88] train-logloss:0.032766 
## [89] train-logloss:0.032383 
## [90] train-logloss:0.031963 
## [91] train-logloss:0.031637 
## [92] train-logloss:0.031239 
## [93] train-logloss:0.030814 
## [94] train-logloss:0.030491 
## [95] train-logloss:0.030148 
## [96] train-logloss:0.029865 
## [97] train-logloss:0.029523 
## [98] train-logloss:0.029290 
## [99] train-logloss:0.029075 
## [100]    train-logloss:0.028823

Predict and Evaluate

xgb_pred <- predict(xgb_model, test_x)
xgb_class <- ifelse(xgb_pred > 0.5, 1, 0)

Confusion Matrix

conf_matrix_xgb <- table(test_y, xgb_class)
conf_matrix_xgb

##       xgb_class
## test_y   0   1
##      0 122  27
##      1  35  46

Accuracy

accuracy_xgb <- sum(diag(conf_matrix_xgb)) / sum(conf_matrix_xgb)
cat("XGBoost Accuracy:", accuracy_xgb)

## XGBoost Accuracy: 0.7304348

AUC

roc_curve_xgb <- roc(test_y, xgb_pred)

## Setting levels: control = 0, case = 1

## Setting direction: controls < cases

auc(roc_curve_xgb)

## Area under the curve: 0.8124

Accuracy comparison

accuracy_table <- data.frame(
  Model = c("Logistic Regression", "Random Forest", "SVM", "XGBoost"),
  Accuracy = c(accuracy, accuracy_rf, accuracy_svm, accuracy_xgb)
)
print(accuracy_table)

##                 Model  Accuracy
## 1 Logistic Regression 0.7913043
## 2       Random Forest 0.7478261
## 3                 SVM 0.8000000
## 4             XGBoost 0.7304348

Plot AUC

plot(roc_curve, col = "red", main = "ROC Curves")
plot(roc_curve_xgb, col = "blue", add = TRUE)
legend("bottomright", legend = c("Logistic", "XGBoost"), col = c("red", "blue"), lwd = 2)

Diabetes Prediction Analysis using Machine Learning Approach in R

Omprakash Sahani

Introduction

Key Achievements

Conclusion

Install Required Packages

Load the installed packages

Load the dataset

Inspect the dataset

Check for missing values

Normalize numerical columns (if needed)

Check for outliers using boxplots

Correlation Matrix

Glucose levels by Outcome

Install and load the caret package

Logistic Regression Model

Predict and Evaluate

Confusion Matrix

Accuracy

AUC

Train Random Forest model

Predict

Confusion Matrix

Accuracy

Feature Importance

SVM Model

Confusion Matrix

Accuracy

Prepare data for XGBoost

XGBoost Model

Predict and Evaluate

Confusion Matrix

Accuracy

AUC

Accuracy comparison

Plot AUC