---
title: "Iris Prediction Model Evaluation"
output:
flexdashboard::flex_dashboard:
orientation: columns
vertical_layout: fill
social: [ "twitter", "facebook","menu" ]
source_code: embed
---
```{r setup, include=FALSE}
library(flexdashboard)
library(MachineShop)
library(survival)
library(MASS)
library(magrittr)
library(gbm)
library(rpart)
library(partykit)
library(randomForest)
library(kknn)
library(caret)
```
```{r include=FALSE}
# attach the iris dataset
data(iris)
# rename the dataset
dataset <- iris
# define the filename
filename <- "iris.csv"
# load the CSV file from the local directory
surv_df <- read.csv(filename, header=FALSE)
# set the column names in the dataset
colnames(surv_df) <- c("Sepal.Length","Sepal.Width","Petal.Length","Petal.Width","Species")
set.seed(123)
# create a list of 80% of the rows in the original dataset we can use for training
validation_index <- createDataPartition(surv_df$Species, p=0.80, list=FALSE)
# select 20% of the data for validation
surv_test <- surv_df[-validation_index,]
# use the remaining 80% of data to training and testing the models
surv_train <- surv_df[validation_index,]
## Global formula for the analysis
surv_fo <- Species ~ .
## All available models
# modelinfo() %>% names
## Model-specific information
# modelinfo(RandomForestModel)
## Control parameters for K-fold cross-validation
## Prediction of survival means
surv_means_control <- CVControl(folds = 5, repeats = 3, seed = 123)
## Prediction of survival probabilities
surv_probs_control <- CVControl(folds = 5, repeats = 3, seed = 123)
```
Random Forest Model
================
Row {.tabset .tabset-fade}
-------------------------
### Prediction of Means and Probability
```{r}
## Model function
surv_fit_RF <- fit(surv_fo, data = surv_train, model = RandomForestModel)
## Observed responses
obs <- response(surv_fo, surv_test)
## Predicted survival means
pred_means_RF <- predict(surv_fit_RF, newdata = surv_test)
performance(obs, pred_means_RF)
## Predicted survival probabilities
pred_probs_RF <- predict(surv_fit_RF, newdata = surv_test, type = "prob")
performance(obs, pred_probs_RF)
(res_probs_RF <- resample(surv_fo, data = surv_df, model = RandomForestModel, control = surv_probs_control))
```
### Performance Metrics
```{r}
## Confusion matrices
(conf_RF <- confusion(res_probs_RF, cutoff = 0.5))
performance(conf_RF, metrics = c("Accuracy" = accuracy,
"Sensitivity" = sensitivity,
"Specificity" = specificity))
```
### Confusion Matrix
```{r}
summary(conf_RF)
```
### Confusion Metrix Plot
```{r}
plot(conf_RF)
```
RPartModel
================
Row {.tabset .tabset-fade}
-------------------------
### Prediction of Means and Probability
```{r}
## Model function
surv_fit_RP <- fit(surv_fo, data = surv_train, model = RPartModel)
## Observed responses
obs <- response(surv_fo, surv_test)
## Predicted survival means
pred_means_RP <- predict(surv_fit_RP, newdata = surv_test)
performance(obs, pred_means_RP)
## Predicted survival probabilities
pred_probs_RP <- predict(surv_fit_RP, newdata = surv_test, type = "prob")
performance(obs, pred_probs_RP)
(res_probs_RP <- resample(surv_fo, data = surv_df, model = RPartModel, control = surv_probs_control))
```
### Performance Metrics
```{r}
## Confusion matrices
(conf_RP <- confusion(res_probs_RP, cutoff = 0.5))
performance(conf_RP, metrics = c("Accuracy" = accuracy,
"Sensitivity" = sensitivity,
"Specificity" = specificity))
```
### Confusion Matrix
```{r}
summary(conf_RP)
```
### Confusion Metrix Plot
```{r}
plot(conf_RP)
```
KNNModel
================
Row {.tabset .tabset-fade}
-------------------------
### Prediction of Means and Probability
```{r}
## Model function
surv_fit_knn <- fit(surv_fo, data = surv_train, model = KNNModel)
## Observed responses
obs <- response(surv_fo, surv_test)
## Predicted survival means
pred_means_knn <- predict(surv_fit_knn, newdata = surv_test)
performance(obs, pred_means_knn)
## Predicted survival probabilities
pred_probs_knn <- predict(surv_fit_knn, newdata = surv_test, type = "prob")
performance(obs, pred_probs_knn)
(res_probs_knn <- resample(surv_fo, data = surv_df, model = KNNModel, control = surv_probs_control))
```
### Performance Metrics
```{r}
## Confusion matrices
(conf_knn <- confusion(res_probs_knn, cutoff = 0.5))
performance(conf_knn, metrics = c("Accuracy" = accuracy,
"Sensitivity" = sensitivity,
"Specificity" = specificity))
```
### Confusion Matrix
```{r}
summary(conf_knn)
```
### Confusion Metrix Plot
```{r}
plot(conf_knn)
```
