Neural.network.practice.Rmd

---
title: "Deep Neural Network Regression"
author: "Lyndsey Umsted"
date: '2022-06-24'
output: 
  html_document:
    toc: true
    number_sections: false
---

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

```{r}
suppressMessages(library(readr))
suppressMessages(library(keras))
suppressMessages(library(DT))
```

<style type = "text/css">
h1 {
    color: #1a2451;
}
h2 {
    color: #ffbd4a;
}
h3 {
    color: #1a2451;
}
</style>


## Introduction
Introducing a densely connected deep neural network with two hidden layers and an output layer

## Data
```{r}
aad <- read.csv("Annual_Amazon_Data.csv")
view(aad)

cutaneous <- aad$Cutaneous.Leishmaniasis
mucosal <- aad$Mucosal.Leishmaniasis
visceral <- aad$Visceral.Leishmaniasis
new_data <- subset(aad, !is.na(cutaneous))
#View(new_data)
library(tidyverse)
library(dplyr)
#names(new_data)

## splitting the data into a before 2014 set and an after 2014 set

early_data <- new_data %>%
  filter(Year < 2014)%>%
  dplyr::select(-c(29:69))
later_data <- new_data %>%
  filter(Year > 2013) %>%
  dplyr::select(-c(29:69))

#names(early_data)

## removing unnecessary variables

early_data <- early_data %>%
  dplyr::select(-c("AvgRad"))
later_data <- later_data %>%
  dplyr::select(-c("StableLights")) %>%
  mutate(later_data$OptTemp_Obs <- as.numeric(later_data$OptTemp_Obs)) %>%
  mutate(later_data$Year <- as.numeric(later_data$Year)) %>%
  mutate(later_data$Population <- as.numeric(later_data$Population))

```


```{r, fig.width = 10, fig.height = 10}
later_data_small <- later_data %>%
  dplyr::select(c("Year", "Population", "Cutaneous.Leishmaniasis", "Dengue_Alb_OptTemp", "Dengue_Aeg_OptTemp", "Chik_Alb_OptTemp", "Chik_Aeg_OptTemp", "Zika_OptTemp", "Malaria_OptTemp", "LST_Day", "LST_Night", "OptTemp_Obs", "NDVI", "EVI", "Precip", "AvgRad", "SWOccurrence")) %>%
  na.omit(later_data_small) 
later_data_t <- later_data_small %>%
  filter(Cutaneous.Leishmaniasis > 0)%>%
  mutate(Cutaneous.Leishmaniasis = (Cutaneous.Leishmaniasis)^(-0.025))
# pairs(later_data_small, labels = c("Year", "Population", "Cutaneous.Leishmaniasis","LST_Day", "LST_Night", "OptTemp_Obs", "NDVI", "EVI", "Precip", "AvgRad", "SWOccurrence"))
# library(ggplot2)
# library(GGally)
# ggpairs(data.frame(Year = later_data_small[,1], Population = later_data_small[,2], Cutaneous.Leishmaniasis = later_data_small[,3], LST_Day = later_data_small[,4], LST_Night = later_data_small[,5], OptTemp_Obs = later_data_small[,6], NDVI = later_data_small[,7], EVI = later_data_small[,8], Precip = later_data_small[,9], StableLights = later_data_small[,10], SWOccurrence = later_data_small[,11]), lower = list(continuous = wrap('points'), alpha = 0.3, size  = 0.2))
```



```{r, fig.width = 10, fig.height = 10}
plot(later_data_t)
```

```{r}
# splitting data into training and test data
set.seed(2)
library(caTools)
split <- sample.split(later_data_t, SplitRatio = 0.7)
split
train <- subset(later_data_t, split = "TRUE")
test <- subset(later_data_t, split = "FALSE")
train
test
```

```{r}
# create the model
model <- lm(later_data_t$Cutaneous.Leishmaniasis ~ .,data = train)
summary(model)

```

```{r}
library(MASS)
boxcox(model, plotit = T, lambda = seq(0,2,len = 100))
```


```{r}
plot(model)
```

```{r}
#prediction
pred <- predict(model, test)
pred
```

```{r}
#install.packages("ggpubr")
library(ggpubr)
ggdensity(pred-later_data_t$Cutaneous.Leishmaniasis) + stat_overlay_normal_density(color = "red", linetype = "dashed")
```

```{r}
# comparing predicted vs actual values
par(mfrow = c(1,2), mgp = c(2,1,0), mar = c(3,3,3,1) + 0.1)
plot(test$Cutaneous.Leishmaniasis, type ="l", lty = 1.8, col = "red")
lines(pred, type ="l", col = "blue")
plot(pred, type ="l", lty = 1.8, col = "blue")
```


```{r}
# Finding Accuracy
rmse <- sqrt(mean(pred-later_data_t$Cutaneous.Leishmaniasis)^2)
rmse
```

```{r}
par(mfrow = c(1,3), mgp = c(2,1,0), mar = c(3,3,3,1) + 0.1)
plot(fitted(model), residuals(model), xlab = "fitted values", ylab = "residuals")
abline(h = 0)
title("Residuals vs Fitted")
qqnorm(residuals(model), ylab = "residuals", main = "QQ Plot of Residuals")
qqline(residuals(model))
plot(model, 4)
```