facebook EDA.Rmd

---
title: "Exploratory Data Analysis of facebook ads"
output: github_document
---

```{r message=FALSE, warning=FALSE}
library(tidyverse, quietly = TRUE); library(readr)
FacebookAds <- read_csv("FacebookAds.csv")

```
Let's have a histogram party and look at some of the fields with numbers and categories.
```{r}
FacebookAds_quant <-
  FacebookAds %>%
  select(AdID, AdSpend, Impressions, Clicks, CreationDate,
         EndDate) %>%
  mutate(CreationDate = lubridate::ymd_hms(CreationDate) %>%
           lubridate::decimal_date(),
         EndDate = lubridate::ymd_hms(EndDate) %>%
           lubridate::decimal_date()) %>%
  gather(stat, value, -AdID)

ggplot(FacebookAds_quant, aes(value)) +
  geom_histogram(bins = 100) +
  facet_wrap(~stat, scales = "free")
```
Spend vs. Clicks vs. Impressons

```{r}
FacebookAds %>%
  ggplot(aes(AdSpend, Impressions, color = log(Clicks), 
             label = str_wrap(str_sub(AdText, end = 120)))) +
  geom_point(size = 1/2, alpha = 1/2) +
  scale_x_continuous(trans = "log10", labels = scales::comma) +
  scale_y_continuous(trans = "log10", labels = scales::comma) +
  ggrepel::geom_text_repel(data = FacebookAds %>% top_n(5, Impressions), size = 2)

```