Decision Tree.Rmd

---
title: "Decision Tree Regression Models"
author: "Yihao Zhao"
date: "12/6/2022"
output: html_document
---

## Import the libraries

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


## Import the database

```{r warning = FALSE}
df <- read_excel("globalterrorismdb_0522dist.xlsx")
head(df)
```


## Data cleaning 

```{r}
# Selected features
features = c('country','latitude','longitude','success','suicide','attacktype1','targtype1','gname','claimed','weaptype1','nkill','nwound','property','ransom')
df = df[features]
str(df)
```

```{r}
selected_gname = df %>%
  filter(gname != 'Unknown') %>% 
  group_by(gname) %>% 
  count() %>% 
  arrange(desc(n)) %>% 
  head(10)
  

selected_gname # focus on these most active 10 terrorist groups
```

```{r}
df2 = df %>% 
  filter(gname %in% selected_gname$gname)
head(df2)
```


```{r}
df2[is.na(df2)] <- 0 # fill all NAs to 0
str(df2)
```

## Create train/test set

```{r}
# create a function to split the train and test data set
create_train_test = function(data, size = 0.8, train = TRUE) {
    n_row = nrow(data)
    total_row = size * n_row
    train_sample = 1: total_row
    if (train == TRUE) {
        return (data[train_sample, ])
    } else {
        return (data[-train_sample, ])
    }
}
```


```{r}
df_train <- create_train_test(df2, 0.8, train = TRUE)
df_test <- create_train_test(df2, 0.8, train = FALSE)

dim(df_train)
```

```{r}
dim(df_test)
```
**Interpretation:** The train dataset has 37386 rows while the test dataset has 9347 rows.


```{r}
# verify if the randomization process is correct
prop.table(table(df_train$gname))
```

```{r}
# verify if the randomization process is correct
prop.table(table(df_test$gname))
```



## Build Decision Tree Regressioin Models

```{r}
tree <- rpart(success ~ ., data = df_train, method = 'class')
rpart.plot(tree, extra = 106)
```

## Measure performance

```{r}
# Function that measures the performance
accuracy <- function(tree) {
  
    pred_test <- predict(tree, df_test, type = "class")

    conf <- table(df_test$success, pred_test)
    
    accuracy <- sum(diag(conf))/sum(conf)
    
    accuracy
}
```

```{r}
accuracy(tree)
```

**Interpretation:** The accuracy is 89.97%. This indicates that these data are sufficient enough and appropriate for decision tree modeling.




```{r}
df$gname = as.factor(df$gname)
fit <- rpart(gname ~ ., data = df_train, method = 'class')
rpart.plot(fit, box.palette=0)
```

```{r}
accuracy(fit)
```

**Interpretation:** The accuracy is 8.39%. This indicates that these data are not sufficient enough and appropriate for decision tree modeling.


```{r}
head(df2)
```