Data Visualization Example: Montreal Airbnb Listings

利用Airbnb的搜尋房源結果繪製圖表，並與地圖結合

Data

利用上週 Web Crawling 的結果作為本週資料來源，主要會用到的變數為 district 、lat 、 lng 、 person 、 price

Data File	Variables	Class	Type
house	district	factor	categorical
	lat	numeric	continuous
	lng	numeric	continuous
	person	factor	categorical
	price	numeric	continuous

library(ggplot2)
## Read In Data
house = read.csv("C:/Users/U430/Desktop/CSX Data Science/wk3/Montreal_Airbnb.csv")
attach(house)
house$title = as.character(house$title)
house$person = as.factor(house$person)

str(house)

## 'data.frame':    54 obs. of  9 variables:
##  $ title   : chr  "Urban Retreat in Amazing Location Loft" "SPACIOUS 2 BDRS - WALK EVERYWHERE - EASY CHECK-IN" "Charming & Cosy Room - Well Located" "Perfect Location Studio Apartment" ...
##  $ district: Factor w/ 16 levels "Cote-des-Neiges",..: 2 5 5 5 5 5 1 15 14 5 ...
##  $ lat     : num  45.5 45.5 45.5 45.5 45.5 ...
##  $ lng     : num  -73.6 -73.6 -73.6 -73.6 -73.6 ...
##  $ person  : Factor w/ 6 levels "1","2","3","4",..: 3 4 4 2 2 1 3 4 2 2 ...
##  $ bedrooms: int  0 2 1 0 0 1 1 0 1 1 ...
##  $ beds    : int  1 2 2 1 1 1 1 2 1 1 ...
##  $ type    : Factor w/ 2 levels "entire_home",..: 1 1 2 1 1 2 2 1 2 2 ...
##  $ price   : int  2870 885 794 947 1893 519 458 1191 1435 1252 ...

Reconstruct Data

因為需要繪製各地區平均價格的圖，因此另外整理了新的 dataframe house2

Data File	Variable	Class	Type
house2	district.1	factor	categorical
	mean.price	numeric	continuous
	max.price	numeric	continuous
	mean.price.type	factor	categorical

mean.price.type為各地區平均價格與全體平均比較之結果

## Reconstruct Data
district.1 = as.character(unique(district))
mean.price = c()
max.price = c()

for(i in district.1) {
    mean.price = c(mean.price, mean(price[district == i])) ## Calculate district mean price
    max.price = c(max.price, max(price[district == i]))    ## Calculate district max price
}

house2 = data.frame(district.1, mean.price, max.price)

### Mean Price Index
for(i in 1:16) {
  if(mean.price[i] > mean(price)) {
    house2$mean.price.type[i] = "Above Average"
  } else {
    house2$mean.price.type[i] = "Below Average"
  } 
 }

str(house2)

## 'data.frame':    16 obs. of  4 variables:
##  $ district.1     : Factor w/ 16 levels "Cote-des-Neiges",..: 2 5 1 15 14 10 13 9 6 8 ...
##  $ mean.price     : num  1744 1280 458 1181 1527 ...
##  $ max.price      : int  2870 2565 458 2260 1832 763 1802 2198 611 1191 ...
##  $ mean.price.type: chr  "Above Average" "Below Average" "Below Average" "Below Average" ...

Graphics

單一類別變數 – Bar Chart

列出各Neighborhood房源數量，並以可居住人數作分類

district.count = 
  ggplot(data = house, aes(x = district)) +
  geom_bar(width = 0.5, aes(fill = person))+
  coord_flip() +
  theme_bw() +
  theme(legend.position = "bottom", 
        axis.text.x = element_text(size = 6)) +
  scale_fill_discrete(name = "Person Capacity",
                      guide = guide_legend(reverse = TRUE)) +
  scale_y_continuous(breaks = seq(1,15,1), name = "Count") +
  scale_x_discrete(name = "") +
  labs(title = "Montreal Airbnb House Listing Number", subtitle = "By Neighborhood")
  
## Print Plot
print(district.count)

類別變數 vs. 連續變數 – Bar Chart

列出各Neighborhood平均房源價格

district.price = 
  ggplot(data = house2, aes(x = district.1, y = mean.price)) +
  geom_bar(stat = "identity", width = 0.5, fill = "skyblue") +
  coord_flip() +
  theme_bw() +
  scale_x_discrete(name = "") +
  scale_y_continuous(name = "Mean Price/Night (CAD)") +
  labs(title = "Montreal Airbnb House Listing Price", subtitle = "Avg. Price By Neighborhood")

## Print Plot
print(district.price)

類別變數 vs. 連續變數 – Point Chart

標出各Neighborhood平均價格點後，並根據是否高於總體平均標註顏色

price.comparison = 
  ggplot(data = house2, aes(x = district.1, y = mean.price, label = round((mean.price), digits = 0))) +
  geom_point(aes(col = mean.price.type), size = 6) +
  coord_flip() +
  theme_bw() +
  scale_color_manual(name = "Mean Price Comparison", 
                     labels = c("Above Average", "Below Average"), 
                     values = c("Above Average" = "#FF9999", "Below Average" = "skyblue")) +
  geom_text(color = "#333333", size = 3) +
  scale_x_discrete(name = "") +
  theme(legend.position = "bottom") +
  geom_hline(yintercept = 1502,linetype = "dashed") +
  annotate("text", y = 1502, x = 1, label = "Ave. Price = $1502", size = 4) +
  scale_y_continuous(name = "Mean Price/Night (CAD)") +
  labs(title = "Montreal Airbnb House Listing Price", subtitle = "Avg. Price By Neighborhood")

## Print Plot
print(price.comparison)

Map Plot

將資料點與OpenStreetMap地圖結合

載入OpenStreetMap地圖圖層

library(OpenStreetMap)
mtl.map = openmap(c(lat = 45.5800, lon = -73.6500), c(lat = 45.4177, lon = -73.5290),
                  type = "osm", zoom = 12)
mtl.map.plot = autoplot(openproj(mtl.map))

於地圖圖層上利用ggplot2語法繪製座標點，並以價格作為標籤

mtl.airbnb.price = mtl.map.plot +
  geom_label(data = house, aes(x = lng, y = lat, label = price, fontface = 3), size = 3, hjust = 1, vjust = 0) +
  geom_point(data = house, aes(x = lng, y = lat, fill = person), size = 2, shape = 21) +
  scale_fill_manual(name = "Person Capacity", values = c("#999999", "#E69F00", "#56B4E9", "#009E73", "#F0E442", "#0072B2")) +
  labs(title = "Montreal Airbnb Listings", x = "Longtitude", y = "Latitude") +
  theme(plot.title = element_text(face = "bold", size = 14))

## Print Plot
print(mtl.airbnb.price)

後記

ggplot2雖有基本的統計運算，但似乎不能直接計算類別平均或是做相關分類，需要另外再整理出表格來製圖
ggmap package的get_googlemap目前都需要API來執行，因此另外選擇了OpenStreetMap，而此地圖開發者也直接製作了R package直接取用，缺點是地圖沒辦法放大縮小，這種有進階功能的圖層還是需要register API，但OpenStreetMap的API可透過註冊免費取得。

Data Visualization: Montreal Airbnb Example

graduatecrisis

Data Visualization Example: Montreal Airbnb Listings

Data

Reconstruct Data

Graphics

單一類別變數 – Bar Chart

類別變數 vs. 連續變數 – Bar Chart

類別變數 vs. 連續變數 – Point Chart

Map Plot

後記