eda.Rmd

```{r setup, include=FALSE} 
knitr::opts_chunk$set(warning = FALSE, message = FALSE)
```
```{r}
library(readr)
library(dplyr)
library(ggplot2)
library(corrplot)
library(gridExtra)
```


```{r}
df <- read_csv("data/heart.csv")
```

```{r}
head(df)
```


```{r}
summary(df)
```

```{r}
paste("NA Values:", sum(is.na(df)))
paste("Duplicate Values:", sum(duplicated(df)))
```



```{r}
### Continuous Attributes
# trtbps
# chol
# thalach
# age


# Viewing the Box Plots of these attributes to determine if outliers exist

par(mfrow=c(2,2))

par(mar=c(1,3,1,1))
values = c("age", "chol", "trtbps", "thalachh")


for(name in values){
        value <- df[name]
        boxplot(value)
        stripchart(value, vertical=TRUE, col="blue", add=TRUE, pch = 20, method="jitter")
}

par(mfrow=c(1,1))

# Based on the results, outliers exist in these attributes of the data
```

```{r}

# Spearman Correlation should be weaker since there are outliers in the data
# Kendall Correlation should be weaker since most of the attributes are not dependent upon each other.

par(mfrow=c(2,2))
corrplot(cor(df, method=c("spearman")), title="Spearman",mar=c(0,0,2,0), type="lower", tl.pos="n", cl.pos="n", method="square", order="FPC")
corrplot(cor(df, method=c("pearson")), title="Pearson", mar=c(0,0,2,0), type="lower", tl.pos="n", cl.pos="n", method="square", order="FPC")
corrplot(cor(df, method=c("kendall")), title="Kendall", mar=c(0,0,2,0), type="lower", tl.pos="n", cl.pos="n", method="square", order="FPC")

par(mfrow=c(1,1))

# As expected, the strength of the correlation in Pearson is slightly better than Spearman and Kendall

```

```{r}
# Spearman is our desired correlation method

corrplot(cor(df, method=c("spearman")), type="lower", method="square", order="FPC")
```

```{r}

# Viewing the Age/Output Distribution
ggplot(data=df, aes(x=age, fill=output, group=output)) +
        geom_histogram(position="stack")


# Ages 30-50 has a high chance of heart attack
# Ages 50-70 had a 0.5 chance on average
```


```{r}

# Viewing the Gender/Output Distribution

ggplot(data = df, aes(x = sex, fill=output, group=output)) +
       geom_histogram(stat="count") +
       ggtitle("Gender Survival Rate Distribution")

# People with sex = 0 had a high chance of heart attack
# People with sex = 1 had a 0.5 chance on average
```


```{r}

# Viewing the (Chest Pain)/Output Distribution
ggplot(data=df, aes(x=cp, fill=output, group=output)) +
        geom_histogram(stat="count")

# Chest Pain Type 1,2,3 had a high chance of heart attack
# Chest Pain Type 0 had a lower than 0.5 chance
```

```{r}

# Viewing the (Resting Electrocardiographic Results)/Output Distribution

ggplot(data=df, aes(x=restecg, fill=output, group = output)) +
        geom_histogram(stat="count")

# Type 1 had a high chance of heart attack
# Type 0 had a 0.5 chance
# Type 2 had a low chance of heart attack
```

```{r}

# Viewing the (Exercise Induced Angina)/Output Distribution

ggplot(data=df, aes(x=exng, fill=output, group = output)) +
        geom_histogram(stat="count")

# Type 0 had around a 75% chance
# Type 1 had around a 25% chance
```

```{r}

# Viewing the (Number of Major Vessels)/Output Distribution

ggplot(data=df, aes(x=caa, fill=output, group = output)) +
        geom_histogram(stat="count")

# Type 1,2,3 had a low chance
# Type 4 had a very high chance
# Type 0 had around a 75% chance
```

```{r}

# Viewing the (Fasting Blood Sugar > 120)/Output Distribution

ggplot(data=df, aes(x=fbs, fill=output, group = output)) +
        geom_histogram(stat="count")

# Both types have a 50% chance
# This attribute is not very useful in classification
```

```{r}

# Viewing the (Resting Blood Pressure)/Output Distribution

ggplot(data=df, aes(x=trtbps, fill=output, group=output)) +
        geom_density(alpha=0.6)

# Mostly equal in terms of values of 1.0 and 0
# At Blood Pressure of higher than 190, the chance of heart attack decreases
```

```{r}

# Viewing the (Maximum Heart Rate Achieved)/Output Distribution

ggplot(data=df, aes(x=thalachh, fill=output, group=output)) +
        geom_density(alpha=0.6)

# Left end of the spectrum had lower chance
# Chance increases drastically at 150
```

```{r}

# Viewing the Output Distribution

ggplot(data=df, aes(x=output)) +
        geom_histogram(position="stack", binwidth=1, color="black", size=2)

# We have slightly more values for 1.0 than 0.0 so we will need to sample properly
```