---
title: "Report on (data set)"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{Template}
  %\VignetteEncoding{UTF-8}
  %\VignetteEngine{knitr::rmarkdown}
editor_options: 
  markdown: 
    wrap: 72
---

# Abstract here

# Introduction

# Statement of the problem from the customer's perspective

# Literature review/summary, history of previous results

# The goal of this investigation

# Exploratory Data Analysis

1.  Head of data frame (put report here)

2.  Data summary (in Console)

3.  Variance Inflation Factor report

4.  Correlation of the data (table)

5.  Histograms of each numeric column

6.  Boxplots of the numeric data

7.  Each feature vs target (by percent)

8.  Each feature vs target (by number)

9.  Correlation plot of the numeric data (as circles and colors)

10. Correlation plot of the numeric data (as numbers and colors)

11. Correlation of the data (report)

# Model building

### Function call (replace with your function call):

```         
library(ClassificationEnsembles)

Classification(data = ISLR::Carseats,
               colnum = 7,
               numresamples = 25,
               predict_on_new_data = "N",
               save_all_plots = "N",
               set_seed = "N",
               how_to_handle_strings = 1,
               remove_VIF_above <- 5.00,
               save_all_trained_models = "N",
               scale_all_numeric_predictors_in_data = "N",
               use_parallel = "N",
               train_amount = 0.50,
               test_amount = 0.25,
               validation_amount = 0.25)
```

Discussion of function call here. (For example, the code above randomly
resamples the data 25 times, and sets train = 0.50, test = 0.25,
validation = 0.25, you might want to discuss other aspects of the
function call. For example, the function call does not set a seed, so
the results are random.)

### **List of models (individual models first):**

**C50:**

```         
C50_train_fit <- C50::C5.0(as.factor(y_train) ~ ., data = train)
```

**Linear:**

```         
linear_train_fit <- MachineShop::fit(y ~ ., data = train01, model = "LMModel")
```

**Partial Least Squares:**

```         
pls_train_fit <- MachineShop::fit(y ~ ., data = train01, model = "PLSModel")
```

**Penalized Discriminant Analysis**

```         
pda_train_fit <- MachineShop::fit(y ~ ., data = train01, model = "PDAModel")
```

**RPart:**

```         
rpart_train_fit <- MachineShop::fit(y ~ ., data = train01, model = "RPartModel")
```

**Trees:**

```         
tree_train_fit <- tree::tree(y_train ~ ., data = train)
```

**How the ensemble is made:**

```         
ensemble1 <- data.frame(
    "C50" = c(C50_test_pred, C50_validation_pred),
    "Linear" = c(linear_test_pred, linear_validation_pred),
    "Partial_Least_Squares" = c(pls_test_pred, pls_validation_pred),
    "Penalized_Discriminant_Analysis" = c(pda_test_pred, pda_validation_pred),
    "RPart" = c(rpart_test_pred, rpart_validation_pred),
    "Trees" = c(tree_test_pred, tree_validation_pred)
  )

ensemble_row_numbers <- as.numeric(row.names(ensemble1))
ensemble1$y <- df[ensemble_row_numbers, "y"]

ensemble1 <- ensemble1[complete.cases(ensemble1), ]
```

**Ensemble Bagged Cart:**

```         
ensemble_bag_cart_train_fit <- ipred::bagging(y ~ ., data = ensemble_train)
```

**Ensemble Bagged Random Forest:**

```         
ensemble_bag_train_rf <- randomForest::randomForest(ensemble_y_train ~ ., data = ensemble_train, mtry = ncol(ensemble_train) - 1)
```

**Ensemble C50:**

```         
ensemble_C50_train_fit <- C50::C5.0(ensemble_y_train ~ ., data = ensemble_train)
```

**Ensemble Naive Bayes:**

```         
ensemble_n_bayes_train_fit <- e1071::naiveBayes(ensemble_y_train ~ ., data = ensemble_train)
```

**Ensemble Support Vector Machines:**

```         
ensemble_svm_train_fit <- e1071::svm(ensemble_y_train ~ ., data = ensemble_train, kernel = "radial", gamma = 1, cost = 1)
```

**Ensemble Trees:**

```         
ensemble_tree_train_fit <- tree::tree(y ~ ., data = ensemble_train)
```

# Model evaluations

1.  Model accuracy (put model accuracy barchart here)

2.  All confusion matrices (in console)

3.  Over or underfitting barchart

4.  True positive rate by model and resample (choose fixed scales or
    free scales)

5.  True negative rate by model and resample (choose fixed or free
    scales)

6.  False positive rate by model and resample (choose fixed or free
    scales)

7.  False negative rate by model and resample (choose fixed or free
    scales)

8.  Duration barchart

9.  Accuracy by model and resampling (chose fixed or free scales)

10. Accuracy data, including train and holdout (choose fixed or free
    scales)

11. Classification error by model and resample (choose fixed or free
    scales)

12. Residuals by model and resample (choose fixed or free scales)

13. Holdout accuracy / train accuracy by model and resample (choose
    fixed or free scales)

14. Head of ensemble (report)

15. Variance Inflation Factor report

# Final Model Selection

1.  Most accurate model:

2.  Mean Holdout Accuracy

3.  Standard deviation of mean holdout accuracy

4.  Classification error mean

5.  Duration (mean)

6.  True positive rate (mean)

7.  True negative rate (mean)

8.  False positive rate (mean)

9.  False negative rate (mean)

10. Positive predictive value (mean)

11. Negative predictive value (mean)

12. Prevalence (mean)

13. Detection rate (mean)

14. Detection prevalence (mean)

15. F1 Score

16. Train accuracy (mean)

17. Test accuracy (mean)

18. Validation accuracy (mean)

19. Holdout vs train (mean)

20. Holdout vs train standard deviation

# Strongest evidence based recommendations with margins of error(s)

# Comparison of current results vs previous results

# Future goals with this data set

# References