library(ggplot2)Warning: package 'ggplot2' was built under R version 4.4.3library(tidyverse)
library(dplyr)
library(maps)
library(gganimate)
library(gifski)
library(transformr)
library(av)
library(tidytext)Exploring Airbnb Of New York City
Import and Combine the datasets
AB_NYC <- read_csv("AB_NYC_2019.csv")
AB_NYC_2023 <- read_csv("AB_NYC_2023.csv")
combined_dataset <- bind_rows(AB_NYC, AB_NYC_2023)
combined_dataset <- within(combined_dataset,
rm(number_of_reviews_ltm, license))Information about the data
dim(AB_NYC)[1] 48895 16dim(AB_NYC_2023)[1] 42931 18dim(combined_dataset)[1] 91826 16Missing Value
sum(is.na(combined_dataset))[1] 40764missing_values <- is.na(combined_dataset$host_name)
sum(missing_values)[1] 26combined_dataset_with_missing_values <- combined_dataset |> filter(if_any(everything(), ~is.na(.)))
dim(combined_dataset_with_missing_values)[1] 20385 16Remove Missing Value
combined_dataset <- combined_dataset[!is.na(combined_dataset$host_name), ]
dim(combined_dataset)[1] 91800 16Duplicates
sum(duplicated(combined_dataset))[1] 11combined_dataset_unique <- distinct(combined_dataset)
dim(combined_dataset_unique)[1] 91789 16Boxplot
ggplot(combined_dataset_unique, aes(x = reorder(room_type, -price), y = price, fill = room_type)) +
scale_fill_manual(values = c("#0060CC","#00FFFF", "#9999FF", "#9933FF")) +
geom_boxplot(outlier.shape = NA, outlier.colour = NA) + # no showing outliers
coord_cartesian(ylim = c(0, max(combined_dataset_unique$price) * 0.006)) +
labs(title = "Boxplots of Price by Room Type", x = "Room Type", y = "Price") +
theme(axis.text.x = element_text(size = 10))
ggplot(combined_dataset_unique, aes(x = reorder(room_type, -minimum_nights), y = minimum_nights, fill = room_type)) +
scale_fill_manual(values = c("#0060CC","#00FFFF", "#9999FF", "#9933FF")) +
geom_boxplot(outlier.shape = NA, outlier.colour = NA) +
coord_cartesian(ylim = c(0, max(combined_dataset_unique$minimum_nights) * 0.06)) +
labs(title = "Boxplots of Minimum Nights by Room Type", x = "Room Type", y = "Minimum Nights") +
theme(axis.text.x = element_text(size = 10))
ggplot(combined_dataset_unique, aes(x = reorder(neighbourhood_group, -price),
y = price, fill = neighbourhood_group)) +
scale_fill_manual(values = c("#0060CC","#00FFFF", "#9999FF", "#9933FF","darkblue")) +
geom_boxplot(outlier.shape = NA, outlier.colour = NA) +
coord_cartesian(ylim = c(0, max(combined_dataset_unique$price) * 0.006)) +
labs(title = "Boxplots of Price by Region", x = "Region", y = "Price") +
theme(axis.text.x = element_text(size = 10))
Barplot
combined_dataset_unique |> group_by(room_type) |>
summarise(count = n()) |>
ggplot(aes(x = reorder(room_type, -count), y = count)) +
geom_bar(stat = "identity", fill = "#99CCCC", color = "black") +
labs(title = "Borplot of number of Room Type", x = "Room Type", y = "Count")
The entire home/department and private room are more popular in Airbnb.
combined_dataset_unique |>
group_by(room_type) |>
summarise(count = n(),
avg_price = mean(price)) |>
ggplot(aes(x = reorder(room_type, -avg_price), y = avg_price)) +
geom_bar(stat = "identity", fill = "#99CCCC", color = "black") +
labs(title = "Borplot of Average Price per Room Type",
x = "Room Type", y = "Average Price")
The hotel room are the most expensive room type compared to others. And the shared room are the cheapest.
histogram
numeric_columns <- sapply(combined_dataset_unique, is.numeric)
par(mfrow = c(2, 5))
par(mar = c(2, 2, 2, 2))
for (col in names(combined_dataset_unique[, numeric_columns]))
{ hist(combined_dataset_unique[[col]],
main = col,
xlab=col,
col= "#99CCCC",
border="black") }
The histograms doesn’t look good.
ggplot(combined_dataset_unique, aes(x = availability_365)) +
geom_histogram(binwidth = 10, fill = "#99CCCC", color = "black") +
labs(title = "Availability Distribution", x = "Availability (in days)", y = "Frequency") +
theme_minimal()
Most of the rooms in airbnb are being booked often.
price_upper_limit <- quantile(combined_dataset_unique$price, 0.95)
price_filtered <- combined_dataset_unique |>
filter(price >= 0, price <= price_upper_limit)
# 95% of the data
ggplot(price_filtered, aes(x = price)) +
geom_histogram(fill = "#99CCCC", color = "black") +
labs(title = "Price Distribution", x = "Price") +
facet_wrap(~room_type) +
theme_minimal()
For the entire home/apartment, we can see the price distribution is slightly skewed to the left. Most of the entire home/apartment cost about the same price per night. Compare to the private room, most of the entire home/apartment cost higher than the private room. There is an obvious skewed to the left distribution for the price of private room per night. Most of the private rooms are cheaper.
nights_upper_limit <- quantile(combined_dataset_unique$minimum_nights, 0.95)
nights_filtered <- combined_dataset_unique |>
filter(minimum_nights >= 0,
minimum_nights <= nights_upper_limit)
# 95% of the data
ggplot(nights_filtered, aes(x = minimum_nights)) +
geom_histogram(fill = "#99CCCC", color = "black") +
labs(title = "Minimum Nights Distribution", x = "Minimum Nights") +
theme_minimal()
There is a short-term rental regulations in NYC airbnb. While waiting for approval, the host may set your calendar to 30 nights minimum so that the OSE may review and approve the listing’s eligibility for short-term rental registration. That is why there are a lot of count at the minimum nights of 30. Most of the rooms with a under 10 minimum days to book.
reviews_upper_limit <- quantile(combined_dataset_unique$number_of_reviews, 0.95)
reviews_filtered <- combined_dataset_unique |>
filter(number_of_reviews >= 0, number_of_reviews <= reviews_upper_limit)
ggplot(reviews_filtered, aes(x = number_of_reviews)) +
geom_histogram(fill = "#99CCCC", color = "black") +
labs(title = "Number of Reviews Distribution", x = "Number of Reviews") +
theme_minimal()
There is a obvious skewed to the right. most of the host recieved less reviews.
NY Map data
states <- map_data("state")
counties <- map_data("county")
NewYork <- subset(states, region == "new york")
head(NewYork) long lat group order region subregion
9017 -73.92874 40.80605 34 9050 new york manhattan
9018 -73.93448 40.78886 34 9051 new york manhattan
9019 -73.95166 40.77741 34 9052 new york manhattan
9020 -73.96312 40.75449 34 9053 new york manhattan
9021 -73.96885 40.73730 34 9054 new york manhattan
9022 -73.97458 40.72584 34 9055 new york manhattanggplot(NewYork, aes(x = long, y = lat, group = group, fill = subregion)) +
geom_polygon() +
scale_fill_manual(values = c("#003333","#336666", "#669999", "#99CCCC")) +
coord_map()
NYC_map <- NewYork |> filter(long >= -74.3 & long <= -71 & lat >= 40.4 & lat <= 41.5)ggplot(NYC_map, aes(x = long, y = lat, group = group, fill = subregion)) + geom_polygon() + coord_map() +
scale_fill_manual(values =c("#003333","#336666", "#669999", "#99CCCC"))
NYC Airbnb map
NYCAB_map <- NewYork |> filter(long >= -74.3 & long <= -73.5 & lat >= 40.4 & lat <= 41)
plot_data <-
combined_dataset |>
select(price, longitude, latitude, room_type, neighbourhood_group) |>
group_by(neighbourhood_group, room_type) |>
summarise(count = n(),
price = mean(price),
longitude = median(longitude),
latitude = median(latitude))`summarise()` has grouped output by 'neighbourhood_group'. You can override
using the `.groups` argument.ggplot(NYCAB_map, aes(x = long, y = lat, group = group)) +
geom_polygon(aes(fill = subregion)) +
scale_fill_manual(values = c("#003333","#336666", "#669999", "#99CCCC")) +
geom_point(plot_data,
mapping = aes(x = longitude, y = latitude,
color = room_type, size = price,
group = NULL)) +
scale_color_manual(values = c("#0060CC","#00FFFF", "#9999FF", "#9933FF")) +
coord_map() +
theme_classic() +
facet_wrap(~room_type)
The color of the dots represents different regions of NYC. The hotel room type are mostly around the Manhattan region. The size of the dots are the price. Larger size represents higher price per night. The biggest dots are located in the Manhattan where cost the most expensive to stay in NYC.
- create an animation for the map
anim_room_type <- ggplot(NYCAB_map, aes(x = long, y = lat, group = group)) +
geom_polygon(aes(fill = subregion)) +
scale_fill_manual(values = c("#003333","#336666", "#669999", "#99CCCC")) +
geom_point(plot_data,
mapping = aes(x = longitude, y = latitude,
color = room_type, size = price,
group = NULL)) +
scale_color_manual(values = c("#0060CC","#00FFFF", "#9999FF", "#9933FF")) +
coord_map() +
theme(panel.background = element_blank(),
legend.key = element_blank()) +
transition_states(states = room_type)
anim_room_type
- Save the animation
# anim_save("anim_room_type.gif", animation = anim_room_type)Time series plot
- Table
combined_dataset |>
filter(number_of_reviews > 0) |>
group_by(Year = year(last_review)) |>
summarise(Last_Reviews = n()) |>
arrange(desc(Last_Reviews))# A tibble: 13 × 2
Year Last_Reviews
<dbl> <int>
1 2019 27439
2 2023 13291
3 2022 8570
4 2018 7478
5 2017 4409
6 2016 3982
7 2015 2154
8 2020 2109
9 2021 1592
10 2014 301
11 2013 79
12 2012 37
13 2011 10This data was collected for the activities of Airbnb in the year 2019 and 2023. However, the dataset contains the date of last review received from 2011 to 2023. The date of the last review in other years may be because no one has rented the room since that year so there are no reviews or the tenant did not leave a review.
From the table, we can see that more reviews of Airbnb rooms have been updated to more recent years. More people keep booking rooms in Airbnb and wrote review about their experiments.
Although, the data in 2023 only contains up to March, the updated reviews in 2023 is almost half of the updated reviews in the whole year of 2019. The reason may because after covid, more people are excited to have fun out of their house.
class(combined_dataset_unique$last_review)[1] "Date"combined_dataset_unique |>
group_by(date = last_review) |>
summarise(count = n(),
Avg_price = mean(price, na.rm = T)) |>
filter(count >= mean(count),
year(date) <= 2023,
year(date) >= 2018) |>
ggplot(aes(x = date, y = count)) +
geom_line(linewidth = 0.3, color = "skyblue") +
coord_cartesian(ylim = c(0, 1500)) +
geom_smooth(se = F, linewidth = 0.5) +
labs(x = "Date",
y = "Count of Reviews",
title = "Time Series Plot") +
theme_classic()
class(combined_dataset_unique$last_review)[1] "Date"combined_dataset_unique |>
group_by(date = last_review) |>
summarise(count = n(),
Avg_price = mean(price, na.rm = T)) |>
filter(count > 150,
year(date) <= 2023,
year(date) >= 2023) |>
ggplot(aes(x = date, y = count)) +
geom_line(linewidth = 0.3, color = "skyblue") +
coord_cartesian(ylim = c(0, 1500)) +
geom_smooth(se = F, linewidth = 0.5) +
labs(x = "Date",
y = "Count of Reviews",
title = "Time Series Plot (2023)") +
theme_classic()
The data in 2023 only contains up to March.
class(combined_dataset_unique$last_review)[1] "Date"combined_dataset_unique |>
group_by(date = last_review) |>
summarise(count = n(),
Avg_price = mean(price, na.rm = T)) |>
filter(count >= mean(count),
year(date) <= 2019,
year(date) >= 2019) |>
ggplot(aes(x = date, y = count)) +
geom_line(linewidth = 0.3, color = "skyblue") +
coord_cartesian(ylim = c(0, 1500)) +
geom_smooth(se = F, linewidth = 0.5) +
labs(x = "Date",
y = "Count of Reviews",
title = "Time Series Plot (2019)") +
theme_classic()
In 2019, we can see there is a peak of date of last review be leaved around July 2019. The reason may because more people went for vacation during the summer holiday and then went back to work/school after that.
Most Frequently used word of description about the room
text_data = combined_dataset |>
unnest_tokens(word, name) |>
anti_join(stop_words) |>
count(word, sort = TRUE)Joining with `by = join_by(word)`text_data# A tibble: 11,820 × 2
word n
<chr> <int>
1 bedroom 16464
2 private 13801
3 apartment 12510
4 cozy 9222
5 apt 8476
6 1 8253
7 studio 7971
8 2 7658
9 brooklyn 7566
10 spacious 6665
# ℹ 11,810 more rowstext_data |>
filter(n > 4000, word != "bedroom", word != "apartment") |>
mutate(word = reorder(word, n)) |>
ggplot(aes(n, word)) +
geom_col() +
ylab(" ") +
xlab("Description Word Freq") +
theme_minimal()
Most commonly used words to describe the listings are: cozy, spacious, and beautiful.