FINAL_NNET_WORKING.Rmd

---
title: "R Notebook"
output: html_notebook
---
```{r}
rm(list = ls())
```

```{r}
#define functions and libraries
library(data.table)
library(neuralnet)
library(pROC)
library(nnet)
library(reshape)
library(chemometrics)
library(devtools)
source_url('https://gist.githubusercontent.com/fawda123/7471137/raw/466c1474d0a505ff044412703516c34f1a4684a5/nnet_plot_update.r')
 
# scalezero <- function(column){
#   columnz <- ((column*100)/max(column))/100
#   return(columnz)
# }
```

```{r}
data <- fread('/Volumes/200GB/Big Data Outputs/final sentiment code/neuralnetinput.csv', sep =',')
```

```{r}
data <- as.data.frame(data)
class(data)

for (i in seq(1,length(data))){
  colnames(data)[i] <- gsub(' ','',colnames(data)[i])
  colnames(data)[i] <- gsub('/','_',colnames(data)[i])
  colnames(data)[i] <- gsub('-','_',colnames(data)[i])
  colnames(data)[i] <- gsub('!','',colnames(data)[i])
}

str(data)
data[,c(13,84)]
```

```{r}
subset(data[,c(13,84)], is.na(ROI_RATING))
```

```{r}
data$ROI_RATING <- as.factor(data$ROI_RATING)
str(data)
```

```{r}
unique(data$ROI_RATING)
length(data$ROI_RATING)
length(na.omit(data$ROI_RATING))
nrow(data)
nrow(na.omit(data))
```

```{r}
data <- data[,c(-30)]
names(data)
```

```{r}
data2 <- data
unique(data2$ROI_RATING)
data2$ROI_RATING <- as.factor(as.character(data2$ROI_RATING))
unique(data2$ROI_RATING)
```



```{r}
names(data2)
```



```{r}
indexes = sample(1:nrow(data2), size=0.2*nrow(data2))
train <- data2[-indexes,]
test<- data2[indexes,]
train.title <- train[,1]
test.title <- test[,1]

train <- train[,c(2:12,14:ncol(train))]
test <- test[,c(2:12,14:ncol(test))]

train$ROI_RATING <- class.ind(train$ROI_RATING)
test$ROI_RATING <- class.ind(test$ROI_RATING)

```

```{r}
unique(train$ROI_RATING)
```

```{r}
#get formula
n <- names(train)
f <- as.formula(paste("ROI_RATING ~", paste(n[!(n %in% "ROI_RATING")], collapse = " + ")))
f
n.input <- train[n[!(n %in% "ROI_RATING")]]
n.output <- as.data.frame(train$ROI_RATING)
class(n.input)
class(n.output)
```

```{r}
# #train the neural network using neuralnet
#nn.neuralnet <- neuralnet(f, data = train, hidden = 40,linear.output = F)
```

```{r}
#nn.nnet <- nnet(n.input, n.output, data = train, size = 40,MaxNWts = 11000)
nn.nnet <- multinom(f, data = train)

```

```{r}
names(nn.nnet)
```
```{r}
plot.nnet(nn.nnet, struct= struct)
```

```{r}
names(test)
```

```{r}
predicted.results <- predict(nn.nnet,test[,c(1:80)])
```

```{r}
predicted.results <- as.data.frame(predicted.results)
```



```{r}
cleanoutput <- test
```

```{r}
names(cleanoutput)
```

```{r}
#colnames(cleanoutput)[145] <- "Expected ROI"
cleanoutput<- cbind(cleanoutput,predicted.results)
names(cleanoutput)
```

```{r}
finalresults <- cbind(test.title,cleanoutput)
```

```{r}
names(finalresults)
```

```{r}
finalresults.end <- finalresults[,c(1,82:83)]
```
```{r}
str(finalresults.end)
```

```{r}
class(finalresults.end)
write.csv(finalresults.end, file = 'nnetresults.csv', na = "NA", row.names = FALSE)
```

```{r}
results <- fread('nnetresults-2.csv',sep =',')
```
```{r}
str(results)
```

```{r}
library(caret)
confusionMatrix(results$predicted.results,results$ROI_RATING)
```