| IRIS data| EDA | Decision Tree | Random Forest | SVM |
library(ggplot2) # Data visualization
library(plotly) # Interactive data visualizations
## Warning: package 'plotly' was built under R version 4.3.3
##
## Attaching package: 'plotly'
## The following object is masked from 'package:ggplot2':
##
## last_plot
## The following object is masked from 'package:stats':
##
## filter
## The following object is masked from 'package:graphics':
##
## layout
library(psych) # Will be used for correlation visualizations
## Warning: package 'psych' was built under R version 4.3.3
##
## Attaching package: 'psych'
## The following objects are masked from 'package:ggplot2':
##
## %+%, alpha
library(rattle) # Graphing decision trees
## Warning: package 'rattle' was built under R version 4.3.3
## Loading required package: tibble
## Loading required package: bitops
## Rattle: A free graphical interface for data science with R.
## Version 5.5.1 Copyright (c) 2006-2021 Togaware Pty Ltd.
## Type 'rattle()' to shake, rattle, and roll your data.
library(caret) # Machine learning
## Warning: package 'caret' was built under R version 4.3.3
## Loading required package: lattice
data("iris")#Load the iris data set
head(iris)
## Sepal.Length Sepal.Width Petal.Length Petal.Width Species
## 1 5.1 3.5 1.4 0.2 setosa
## 2 4.9 3.0 1.4 0.2 setosa
## 3 4.7 3.2 1.3 0.2 setosa
## 4 4.6 3.1 1.5 0.2 setosa
## 5 5.0 3.6 1.4 0.2 setosa
## 6 5.4 3.9 1.7 0.4 setosa
summary(iris)
## Sepal.Length Sepal.Width Petal.Length Petal.Width
## Min. :4.300 Min. :2.000 Min. :1.000 Min. :0.100
## 1st Qu.:5.100 1st Qu.:2.800 1st Qu.:1.600 1st Qu.:0.300
## Median :5.800 Median :3.000 Median :4.350 Median :1.300
## Mean :5.843 Mean :3.057 Mean :3.758 Mean :1.199
## 3rd Qu.:6.400 3rd Qu.:3.300 3rd Qu.:5.100 3rd Qu.:1.800
## Max. :7.900 Max. :4.400 Max. :6.900 Max. :2.500
## Species
## setosa :50
## versicolor:50
## virginica :50
##
##
##
## Exploratory data analysis#########
#pairs.panels() from psych package
pairs.panels(
iris[,1:4], # Our data.
scale = TRUE, # Changes size of correlation value lables based on strength.
hist.col = 'grey85', # Histogram color.
bg = c("mediumseagreen","orange2","mediumpurple1")[iris$Species], # Colors of the Species levels.
pch = 21, # The plot characters shape and size.
main = 'Correlation matrix of Iris data') # Title.
#3Dplot
library(plotly)
library(magrittr)
plot_ly(data = iris, # Data
x = ~Sepal.Length, y = ~Petal.Length, z = ~Petal.Width, # X, Y, and Z variables
color = ~Species, # Color separation by Species
type = "scatter3d", # 3D scatterplot
mode = "markers" # Use markers
) %>%
layout(scene = list(xaxis = list(title = 'Sepal length'), # Axes names
yaxis = list(title = 'Petal length'),
zaxis = list(title = 'Petal width')))
#####Boxplot##########
ggplot(
# (1) Set data; (2) Specify X and Y variables; (3) 'fill' color separates our Species levels.
data = iris, mapping = aes(x = Species, y = Sepal.Width, fill = Species)) +
geom_boxplot() + # Specifies that we want a box plot.
scale_fill_brewer(palette = 'Dark2') + # Change color of box plots.
theme_light() + # Set light theme.
labs(title = 'Box plot of sepal width for each species',
x = 'Species', y = 'Sepal width') # Assign a title, axis names.
ggplot(data = iris, mapping = aes(x = Species, y = Sepal.Length, fill = Species)) +
geom_boxplot() +
scale_fill_brewer(palette = 'Dark2') +
theme_light() +
labs(title = 'Box plot of sepal length for each species',
x = 'Species', y = 'Sepal length')
ggplot(data = iris, mapping = aes(x = Species, y = Petal.Width, fill = Species)) +
geom_boxplot() +
scale_fill_brewer(palette = 'Dark2') +
theme_light() +
labs(title = 'Box plot of petal width for each species',
x = 'Species', y = 'Petal width')
ggplot(data = iris, mapping = aes(x = Species, y = Petal.Length, fill = Species)) +
geom_boxplot() +
scale_fill_brewer(palette = 'Dark2') +
theme_light() +
labs(title = 'Box plot of petal length for each species',
x = 'Species', y = 'Petal length')
####Split Data In to Train and Test #####
set.seed(222)
train_index <- createDataPartition(y = iris$Species, # y = our dependent variable.
p = .7, # Specifies split into 70% & 30%.
list = FALSE, # Sets results to matrix form.
times = 1) # Sets number of partitions to create to 1.
train_data <- iris[train_index,] # Use train_index of iris data to create train_data.
test_data <- iris[-train_index,] # Use whatever that is not in train_index to create test_data.
#to predict which category of species (setosa, versicolor, virginica) each iris flower belongs to
####DECISION TREE####
#Model the decision tree model with a 10 fold cross validation.
fitControl <- trainControl(method = "cv", number = 10, savePredictions = TRUE)
#Create a predictor model with the train() function from the CARET package. Specify method = 'rpart' to run a decision tree model.
# Create model
dt_model <- train(Species ~ ., # Set Y variable followed by '~'. The period indicates to include all variables for prediction.
data = train_data, # Data
method = 'rpart', # Specify SVM model
trControl = fitControl) # Use cross validation
confusionMatrix(dt_model)
## Cross-Validated (10 fold) Confusion Matrix
##
## (entries are percentual average cell counts across resamples)
##
## Reference
## Prediction setosa versicolor virginica
## setosa 33.3 0.0 0.0
## versicolor 0.0 29.5 5.7
## virginica 0.0 3.8 27.6
##
## Accuracy (average) : 0.9048
pred_dt <- predict(dt_model, test_data)
# Evaluate accuracy
accuracy_dt <- mean(pred_dt == test_data$Species)
#average accuracy is 90.48% when testing data on resamples of training data. check weather predicted correctly/incorrectly.
#Check the importance of each feature in our model.
# Create object of importance of our variables
dt_importance <- varImp(dt_model)
# Create plot of importance of variables
ggplot(data = dt_importance, mapping = aes(x = dt_importance[,1])) + # Data & mapping
geom_boxplot() + # Create box plot
labs(title = "Variable importance: Decision tree model") + # Title
theme_light() # Theme
#decision tree using fancyRpartPlot() from the RATTLE package. This will give us clear insight into how the model makes its predictions.
fancyRpartPlot(dt_model$finalModel, sub = '')
#PREDICTION: Decision tree model
#Use the created dt_model to run a prediction on the test data.
prediction_dt <- predict(dt_model, test_data)
table(prediction_dt, test_data$Species) %>% # Create prediction table.
prop.table() %>% # Convert table values into proportions instead of counts.
round(2) # Round numbers to 2 significant values.
##
## prediction_dt setosa versicolor virginica
## setosa 0.33 0.00 0.00
## versicolor 0.00 0.31 0.00
## virginica 0.00 0.02 0.33
#Create an object for a 10 fold cross validation. We will use this in our train() function to set trControl next.
fitControl <- trainControl(method = "cv", number = 10, savePredictions = TRUE)
######Random Forest########
# Create model
rf_model <- train(
Species ~ ., # Set Y variable followed by "~." to include all variables in formula.
method = 'rf', # Set method as random forest.
trControl = fitControl, # Set cross validation settings
data = train_data) # Set data as train_data.
# Create object of importance of our variables
rf_importance <- varImp(rf_model)
# Create box plot of importance of variables
ggplot(data = rf_importance, mapping = aes(x = rf_importance[,1])) + # Data & mapping
geom_boxplot() + # Create box plot
labs(title = "Variable importance: Random forest model") + # Title
theme_light() # Theme
confusionMatrix(rf_model)
## Cross-Validated (10 fold) Confusion Matrix
##
## (entries are percentual average cell counts across resamples)
##
## Reference
## Prediction setosa versicolor virginica
## setosa 33.3 0.0 0.0
## versicolor 0.0 29.5 2.9
## virginica 0.0 3.8 30.5
##
## Accuracy (average) : 0.9333
#Prediction: Random forest model
#We will now use our created random forest model in order to predict species on our test data (i.e., the data set our ‘machine’ has not seen before).
#Use the created rf_model to run a prediction on the test data.
prediction_rf <- predict(rf_model, test_data)
table(prediction_rf, test_data$Species) %>% # Create prediction table.
prop.table() %>% # Convert table values into proportions instead of counts.
round(2) # Round numbers to 2 significant values.
##
## prediction_rf setosa versicolor virginica
## setosa 0.33 0.00 0.00
## versicolor 0.00 0.33 0.00
## virginica 0.00 0.00 0.33
pred_rf <- predict(rf_model, test_data)
# Evaluate accuracy
accuracy_rf <- mean(pred_rf == test_data$Species)
#####SVM##########
fitControl <- trainControl(method = "cv", number = 10, savePredictions = TRUE)
# Create model
svm_model <- train(Species ~ ., # Set Y variable followed by '~'. The period indicates to include all variables for prediction.
data = train_data, # Data
method = 'svmLinear', # Specify SVM model
trControl = fitControl) # Use cross validation
confusionMatrix(svm_model)
## Cross-Validated (10 fold) Confusion Matrix
##
## (entries are percentual average cell counts across resamples)
##
## Reference
## Prediction setosa versicolor virginica
## setosa 33.3 0.0 0.0
## versicolor 0.0 30.5 1.0
## virginica 0.0 2.9 32.4
##
## Accuracy (average) : 0.9619
# Create object of importance of our variables
svm_importance <- varImp(svm_model)
# Create box plot
ggplot(data = svm_importance, mapping = aes(x = svm_importance[,1])) + # Data & mapping
geom_boxplot() + # Create box plot
labs(title = "Variable importance: Support vector machine model") + # Title
theme_light() # Theme
prediction_svm <- predict(svm_model, test_data)
table(prediction_svm, test_data$Species) %>% # Create prediction table.
prop.table() %>% # Convert table values into proportions instead of counts.
round(2) # Round numbers to 2 significant values.
##
## prediction_svm setosa versicolor virginica
## setosa 0.33 0.00 0.00
## versicolor 0.00 0.33 0.00
## virginica 0.00 0.00 0.33
pred_SVM <- predict(svm_model, test_data)
# Evaluate accuracy
accuracy_SVM <- mean(pred_SVM == test_data$Species)
# Compare results
results <- data.frame(
Model = c("Decision Tree", "Random Forest", "SVM"),
Accuracy = c(accuracy_dt, accuracy_rf, accuracy_SVM)
)
results
## Model Accuracy
## 1 Decision Tree 0.9777778
## 2 Random Forest 1.0000000
## 3 SVM 1.0000000