A_simple_kNN_model.R

#libraries to create clustering plot of sepal.length and sepal.width of iris dataset
library(ggplot2)
library(dplyr

# Load the Iris dataset
iris_dataset

# Filter the data for setosa, virginica, and versicolor
iris_subset <- iris_dataset %>% filter(Species %in% c("setosa", "virginica", "versicolor"))

# Create a scatter plot with different colors for each class
ggplot(iris_subset, aes(x = Sepal.Length, y = Sepal.Width, color = Species)) +
  geom_point(size = 3) +
  labs(title = "Clustering Plot of Setosa, Virginica, and Versicolor",
       x = "Sepal Length", y = "Sepal Width") +
  theme_minimal()
######################################################################################
#Library to load kNN and CrossTable model evaluation  
install.packages("gmodels")
install.packages("class")
library(gmodels)
library(class)

#load iris dataset
iris_dataset <- iris

#Noarmalize all the numeric columns from 1:4
iris_n <- scale(iris_dataset[,1:4])

#Summary of the idividiual column in iris dataset
summary(iris_n[c("Sepal.Length", "Sepal.Width", "Petal.Length", "Petal.Width")])

# Create train and test dataset. In this case model, I took 80% as train dataset and 20% as test dataset
iris_train <- iris_n[1:120, ]
iris_test <- iris_n[121:150, ]

#Create the labels data frames which contains the species of flowers both for train and test dataset  
iris_train_labels <- iris_dataset[1:120, 5]
iris_test_labels <- iris_dataset[121:150, 5]

#Assess the percentage of each unique labels in the data set
round(prop.table(table(iris_dataset$Species)) * 100, digits = 1)

#run the k-nearest neighbor model to predict test labels of test dataset.   
p <- knn(train = iris_train, test = iris_test, cl = iris_train_labels, k=11)
p

#The test dataset that will be used to estimate the predictive accuracy of the model. 
#Evaluate how well the predicted classes in the iris_test_labels vector match up 
#with the known values in the kNN predicted labels
CrossTable(x = iris_test_labels, y = p, prop.chisq=FALSE)