FeatureEng_GermanCredit.Rmd

---
title: "Feature Engineering German Credit"
output: 
  html_document:
      toc: yes
      toc_float: yes
      code_folding: hide
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```


```{r}
library(dplyr)
library(caret)
library(rpart)
library(rpart.plot)
```


```{r}
# Helper function to print the confusion matrix and other performance metrics of the models.
printPerformance = function(pred, actual, positive="Yes") {
  print(caret::confusionMatrix(data=pred, reference=actual, positive=positive, dnn=c("Predicted", "Actual")))
}
```

# Load Data

```{r}
data(GermanCredit, package = "caret")
df = GermanCredit
df$Class = as.character(df$Class)
df$Class[df$Class == "Bad"] = "NotGood" 
df$Class = as.factor(df$Class)
df.orig = df
str(df)
head(df)
table(df$Class)
```

# Feature Engineering

```{r}
p1 <- preProcess(df['Amount'], 'center')
p2 <- preProcess(df['Amount'], c('center', 'scale'))
p3 <- preProcess(df['Amount'], 'range')
p4 <- preProcess(df['Amount'], 'YeoJohnson')
p5 <- preProcess(df['Amount'], 'BoxCox')

df['Amount.center'] = predict(p1, df['Amount'])
df['Amount.standardization'] = predict(p2, df['Amount'])
df['Amount.range'] = predict(p3, df['Amount'])
df['Amount.YeoJohnson'] = predict(p4, df['Amount'])
df['Amount.BoxCox'] = predict(p5, df['Amount'])

iwidth = 5
iheight = 3

df %>% ggplot(aes(x=Amount)) + 
 geom_histogram(colour="black", fill="white") + 
 xlab("Amount (original)")

ggsave(file="out/gc_amount_1.png", width=iwidth, height=iheight)

df %>% ggplot(aes(x=Amount.range)) + 
 geom_histogram(colour="black", fill="white") + 
 xlab("Amount (range)")

ggsave(file="out/gc_amount_2.png", width=iwidth, height=iheight)

df %>% ggplot(aes(x=Amount.center)) + 
 geom_histogram(colour="black", fill="white") + 
 xlab("Amount (center)")

ggsave(file="out/gc_amount_3.png", width=iwidth, height=iheight)

df %>% ggplot(aes(x=Amount.standardization)) + 
 geom_histogram(colour="black", fill="white") + 
 xlab("Amount (standarization)")

ggsave(file="out/gc_amount_4.png", width=iwidth, height=iheight)

df %>% ggplot(aes(x=Amount.YeoJohnson)) + 
 geom_histogram(colour="black", fill="white") + 
 xlab("Amount (YeoJohnson)")

ggsave(file="out/gc_amount_5.png", width=iwidth, height=iheight)

df %>% ggplot(aes(x=Amount.BoxCox)) + 
 geom_histogram(colour="black", fill="white") + 
 xlab("Amount (BoxCox)")

ggsave(file="out/gc_amount_6.png", width=iwidth, height=iheight)

```

# Splitting the Data

```{r}
df = df.orig
set.seed(123) # Set the seed to make it reproducible

train.index <- createDataPartition(df$Class, p = .8, list = FALSE)
train <- df[ train.index,]
test  <- df[-train.index,]

# Double check that the stratefied sampling worked
table(df$Class)/nrow(df)
table(train$Class)/nrow(train)
table(test$Class)/nrow(test)

actual = test$Class
formula = Class ~ .
positive = "Good"
```



# Evaluation

```{r}
# Function to show the confusion matrix and resulting tree
showResults = function(model){
  pred = predict(model, test)
  print(caret::confusionMatrix(data=pred, reference=actual, positive=positive, dnn=c("Predicted", "Actual")))
  #rpart.plot(model$finalModel, extra=2, type=2)
}
```

```{r}
showResults(weight_fit)
```


# Parameter Tuning - KNN

```{r}
grid <- expand.grid(.kmax = c(5, 10, 25), 
                    .distance=c(1, 2), 
                    .kernel=c("rectangular", "triangular", 
                              "biweight", "cos", "gaussian", "optimal"))

ctrl <- trainControl(method = "repeatedcv", 
                     number = 10, repeats = 5, 
                      classProbs = TRUE, returnResamp = "all")

kknn_fit <- train(formula, 
                  data = train, 
                  method = "kknn", 
                  metric="Kappa", 
                  trControl=ctrl, tuneGrid = grid)

summary(kknn_fit)
kknn_fit
plot(kknn_fit)
showResults(kknn_fit)
```

# Feature Selection

```{r}
# SBF = Selection By Filtering
set.seed(10)
df = df.orig
dim(df)
filterCtrl <- sbfControl(functions = rfSBF)
r <- sbf(formula, data = df, sbfControl = filterCtrl)
r
```

# Model Selection

```{r}
set.seed(123)

ctrl <- trainControl(method = "repeatedcv", 
                     number = 10, repeats = 5, classProbs = TRUE)

rpartFit <- train(formula, data = train, "rpart",
                  preProc=c('nzv', 'center', 'scale'),
                  trControl = ctrl, tuneLength = 10, metric="Kappa")

svmFit <- train(formula, data = train, "svmLinear3",
                  preProc=c('nzv', 'center', 'scale'),
                  trControl = ctrl, tuneLength = 10, metric="Kappa")

nbFit <- train(formula, data = train, "naive_bayes",
                  preProc=c('nzv', 'center', 'scale'),
                  trControl = ctrl, tuneLength = 10, metric="Kappa")

rfFit <- train(formula, data = train, "parRF",
                  preProc=c('nzv', 'center', 'scale'),
                  trControl = ctrl, tuneLength = 10, metric="Kappa")

gbmFit <- train(formula, data = train, "gbm", 
                preProc=c('nzv', 'center', 'scale'),
                trControl = ctrl, tuneLength = 10, metric="Kappa", verbose=F)

kknnFit <- train(formula, data = train, "kknn", 
                 preProc=c('nzv', 'center', 'scale'),
                 trControl = ctrl, tuneLength = 10, metric="Kappa")

# See how things are looking
rpartFit
svmFit
nbFit
rfFit
gbmFit
kknnFit

resamps <- resamples(list(rpart = rpartFit,
                          svm = svmFit,
                          nb = nbFit,
                          rf = rfFit,
                          gbm = gbmFit,
                          kknn = kknnFit))


summary(resamps)
bwplot(resamps)
```