Importing Required Packages

packages = c('rgdal', 'maptools', 'raster','spatstat', 'tmap', 'sf', 'tidyverse','rosm','tmaptools','leaflet','mapview')
for (p in packages){
if(!require(p, character.only = T)){
install.packages(p)
}
library(p,character.only = T)
}

## Loading required package: rgdal

## Loading required package: sp

## rgdal: version: 1.4-8, (SVN revision 845)
##  Geospatial Data Abstraction Library extensions to R successfully loaded
##  Loaded GDAL runtime: GDAL 2.4.2, released 2019/06/28
##  Path to GDAL shared files: /Library/Frameworks/R.framework/Versions/3.6/Resources/library/rgdal/gdal
##  GDAL binary built with GEOS: FALSE 
##  Loaded PROJ.4 runtime: Rel. 5.2.0, September 15th, 2018, [PJ_VERSION: 520]
##  Path to PROJ.4 shared files: /Library/Frameworks/R.framework/Versions/3.6/Resources/library/rgdal/proj
##  Linking to sp version: 1.3-2

## Loading required package: maptools

## Checking rgeos availability: TRUE

## Loading required package: raster

## Loading required package: spatstat

## Loading required package: spatstat.data

## Loading required package: nlme

## 
## Attaching package: 'nlme'

## The following object is masked from 'package:raster':
## 
##     getData

## Loading required package: rpart

## 
## spatstat 1.63-3       (nickname: 'Wet paint') 
## For an introduction to spatstat, type 'beginner'

## 
## Note: spatstat version 1.63-3 is out of date by more than 11 weeks; a newer version should be available.

## 
## Attaching package: 'spatstat'

## The following objects are masked from 'package:raster':
## 
##     area, rotate, shift

## Loading required package: tmap

## Loading required package: sf

## Linking to GEOS 3.7.2, GDAL 2.4.2, PROJ 5.2.0

## Loading required package: tidyverse

## Registered S3 method overwritten by 'cli':
##   method     from    
##   print.boxx spatstat

## ── Attaching packages ────────────────────────────────────────────────────────── tidyverse 1.3.0 ──

## ✓ ggplot2 3.3.0     ✓ purrr   0.3.4
## ✓ tibble  3.0.1     ✓ dplyr   0.8.5
## ✓ tidyr   1.0.2     ✓ stringr 1.4.0
## ✓ readr   1.3.1     ✓ forcats 0.5.0

## ── Conflicts ───────────────────────────────────────────────────────────── tidyverse_conflicts() ──
## x dplyr::collapse() masks nlme::collapse()
## x tidyr::extract()  masks raster::extract()
## x dplyr::filter()   masks stats::filter()
## x dplyr::lag()      masks stats::lag()
## x dplyr::select()   masks raster::select()

## Loading required package: rosm

## Loading required package: tmaptools

## Loading required package: leaflet

## Loading required package: mapview

Importing Data

airbnb <- read_csv("data/aspatial/listings.csv")

## Parsed with column specification:
## cols(
##   id = col_double(),
##   name = col_character(),
##   host_id = col_double(),
##   host_name = col_character(),
##   neighbourhood_group = col_character(),
##   neighbourhood = col_character(),
##   latitude = col_double(),
##   longitude = col_double(),
##   room_type = col_character(),
##   price = col_double(),
##   minimum_nights = col_double(),
##   number_of_reviews = col_double(),
##   last_review = col_date(format = ""),
##   reviews_per_month = col_double(),
##   calculated_host_listings_count = col_double(),
##   availability_365 = col_double()
## )

singapore <- readOGR(dsn = "data/geospatial", layer = "MP14_REGION_WEB_PL")

## OGR data source with driver: ESRI Shapefile 
## Source: "/Users/Amey/Desktop/Y2S2/IS415/assign2/data/geospatial", layer: "MP14_REGION_WEB_PL"
## with 5 features
## It has 9 fields

mpsz <- st_read(dsn = "data/geospatial/master-plan-2014-subzone-boundary-no-sea-shp", layer = "MP14_SUBZONE_NO_SEA_PL")

## Reading layer `MP14_SUBZONE_NO_SEA_PL' from data source `/Users/Amey/Desktop/Y2S2/IS415/assign2/data/geospatial/master-plan-2014-subzone-boundary-no-sea-shp' using driver `ESRI Shapefile'
## Simple feature collection with 323 features and 15 fields
## geometry type:  MULTIPOLYGON
## dimension:      XY
## bbox:           xmin: 2667.538 ymin: 15748.72 xmax: 56396.44 ymax: 50256.33
## proj4string:    +proj=tmerc +lat_0=1.366666666666667 +lon_0=103.8333333333333 +k=1 +x_0=28001.642 +y_0=38744.572 +datum=WGS84 +units=m +no_defs

Converting airbnb to 3414 crs

airbnb4326<-st_as_sf(airbnb,coords = c("longitude","latitude"),crs=4326)
airbnb3414<-st_transform(airbnb4326,3414)
crs(airbnb)

## [1] NA

Section A: Nation Wide Analysis of Airbnb Listings

Viewing all the locations of airbnb listings by house type

The code chunk below creates a map consisting of airbnb listings by house type with openstreemap as its basemap. Therefore, it is created in the view mode. For publishing purposes, a screenshot of the map is attached below it.

tmap_mode("view")
tm_shape(airbnb3414)+
  tm_dots(col = "room_type",
           size = 0.05,
           border.col = "black",
           border.lwd = 1,
          alpha=0.7,
          title = "Room type")+
  tm_basemap(group = "OpenStreetMap")+
  tm_layout(main.title = "Distribution of Airbnb listings by Room Type in Singapore")
tmap_mode("plot")

The map above shows all the Airbnb listings by Room type in Singapore. As seen above, private room, the listing which is indicated in purple is the most popular room type as it can be seen consistently throughout the country in all the regions. Entire home/apartment, which is indicated in green is the second most popular Room type which is also seen in all parts of Singapore, however, most of them are concentrated in central region of Singapore. Shared view is the third most popular room type, and has a few locations distributed among all the regions in Singapore. Lastly, Hotel room is the least popular room type in airbnb and can be found only in the central region of Singapore. To gain deeper understanding and differences in the Room types, we will perform statistcal testing on the data in the next section.

Performing First Order Spatial Analysis

To understand the distribution of the different room types in Singapore, first order spatial analysis is performed on each of the room types to understand if the observations vary from place to place. A distance based approach is preferred over a density based approach due to the fact that there are many regions in Singapore which are unhabitable, such as Central Catchement Area and the Western Catchement Area as they consist of water bodies and vegetation. A nearest neighbour analysis will be performed through the Clark and Evans test which will help us determine if patial randomness exists in the distribution of every room type.

Before the tests are conducted, the data is prepared for every room type. This involves transforming the data from sf to ppp format so that the Clark Evans Test can be performed. Furthermore, there is a check for every room type if duplicated data exists as it is very likely for properties from the same building will have the same coordinates, hence it must be jittered, i.e. shifted minimally so that duplicated data can be eradicated.

Preparing data for each room type

sg_owin <- as(singapore, "owin")

airbnb_spatial <- as(airbnb3414, "Spatial")
airbnb_sp <- as(airbnb_spatial, "SpatialPoints")
airbnb_ppp <- as(airbnb_sp,"ppp")

home <- airbnb3414 %>%
  filter(room_type=="Entire home/apt")
home_spatial <- as(home, "Spatial")
home_sp <- as(home_spatial, "SpatialPoints")
home_ppp <- as(home_sp,"ppp")

hotel <- airbnb3414 %>%
  filter(room_type=="Hotel room")
hotel_spatial <- as(hotel, "Spatial")
hotel_sp <- as(hotel_spatial, "SpatialPoints")
hotel_ppp <- as(hotel_sp,"ppp")

shared <- airbnb3414 %>%
  filter(room_type=="Shared room")
shared_spatial <- as(shared, "Spatial")
shared_sp <- as(shared_spatial, "SpatialPoints")
shared_ppp <- as(shared_sp,"ppp")

private <- airbnb3414 %>%
  filter(room_type=="Private room")
private_spatial <- as(private, "Spatial")
private_sp <- as(private_spatial, "SpatialPoints")
private_ppp <- as(private_sp,"ppp")

home_ppp <- home_ppp[sg_owin]
private_ppp<-private_ppp[sg_owin]
hotel_ppp <-hotel_ppp[sg_owin]
shared_ppp <-shared_ppp[sg_owin]

Checking for duplicated data

To reduce jittering the data, we will check for duplicates in every room type and jitter them individually. The code below does that.

any(duplicated(home_ppp))

## [1] TRUE

sum(multiplicity(home_ppp) > 1)

## [1] 35

sum(multiplicity(hotel_ppp) > 1)

## [1] 59

sum(multiplicity(shared_ppp) > 1)

## [1] 16

sum(multiplicity(private_ppp) > 1)

## [1] 120

home_ppp_jit <- rjitter(home_ppp, retry=TRUE, nsim=1, drop=TRUE)
hotel_ppp_jit <- rjitter(hotel_ppp, retry=TRUE, nsim=1, drop=TRUE)
shared_ppp_jit <- rjitter(shared_ppp, retry=TRUE, nsim=1, drop=TRUE)
private_ppp_jit <- rjitter(private_ppp, retry=TRUE, nsim=1, drop=TRUE)

sum(multiplicity(home_ppp_jit) > 1)

## [1] 0

sum(multiplicity(hotel_ppp_jit) > 1)

## [1] 0

sum(multiplicity(shared_ppp_jit) > 1)

## [1] 0

sum(multiplicity(private_ppp_jit) > 1)

## [1] 0

Plotting the data

To ensure that the converted correctly, we will plot each of the room types individually below.

par(mfrow=c(2,2)) 
plot(sg_owin, main="Private home/apartments listed on Airbnb")
plot(home_ppp_jit, add=TRUE)
plot(sg_owin, main="Hotel rooms listed on Airbnb")
plot(hotel_ppp_jit, add=TRUE)
plot(sg_owin, main="Shared rooms listed on Airbnb")
plot(shared_ppp_jit, add=TRUE)
plot(sg_owin, main="Private rooms listed on Airbnb")
plot(private_ppp_jit, add=TRUE)

Clark and evans test for Entire home/Apartment

The test hypotheses are:

Null Hypothesis: The distribution of entire home/apartment on Airbnb in Singapore are randomly distributed.

Alternate Hypothesis: The distribution of entire home/apartment on Airbnb in Singapore are not randomly distributed.

The 95% confidence interval will be used.

set.seed(1234)
clarkevans.test(home_ppp_jit,
                correction="none",
                clipregion="sg_owin",
                alternative=c("two.sided"),
                nsim=99)

## 
##  Clark-Evans test
##  No edge correction
##  Monte Carlo test based on 99 simulations of CSR with fixed n
## 
## data:  home_ppp_jit
## R = 0.25896, p-value = 0.02
## alternative hypothesis: two-sided

The p-value (0.002) is less than 0.05, hence we can reject the null hypothesis and conclude that we are 95% confident that the distribution of entire home/apartment on Airbnb in Singapore are not randomly distributed. Furthermore, as the R value (0.25) is less than one, there is a significant pattern of clustering among entire home/apartment on Airbnb in Singapore.

Clark and evans test for Hotel rooms

The test hypotheses are:

Null Hypothesis: The distribution of hotel rooms on Airbnb in Singapore are randomly distributed.

Alternate Hypothesis: The distribution of hotel rooms on Airbnb in Singapore are not randomly distributed.

The 95% confidence interval will be used.

clarkevans.test(hotel_ppp_jit,
                correction="none",
                clipregion="sg_owin",
                alternative=c("two.sided"),
                nsim=99)

## 
##  Clark-Evans test
##  No edge correction
##  Monte Carlo test based on 99 simulations of CSR with fixed n
## 
## data:  hotel_ppp_jit
## R = 0.10682, p-value = 0.02
## alternative hypothesis: two-sided

The p-value (0.002) is less than 0.05, hence we can reject the null hypothesis and conclude that we are 95% confident that the distribution of hotel rooms on Airbnb in Singapore are not randomly distributed. Furthermore, as the R value (0.20) is less than one, there is a significant pattern of clustering among hotel rooms on Airbnb in Singapore.

Clark and evans test for Private rooms

The test hypotheses are:

Null Hypothesis: The distribution of private rooms on Airbnb in Singapore are randomly distributed.

Alternate Hypothesis: The distribution of private rooms on Airbnb in Singapore are not randomly distributed.

The 95% confidence interval will be used.

clarkevans.test(private_ppp_jit,
                correction="none",
                clipregion="sg_owin",
                alternative=c("two.sided"),
                nsim=99)

## 
##  Clark-Evans test
##  No edge correction
##  Monte Carlo test based on 99 simulations of CSR with fixed n
## 
## data:  private_ppp_jit
## R = 0.37378, p-value = 0.02
## alternative hypothesis: two-sided

The p-value (0.002) is less than 0.05, hence we can reject the null hypothesis and conclude that we are 95% confident that the distribution of private rooms on Airbnb in Singapore are not randomly distributed. Furthermore, as the R value (0.39) is less than one, there is a significant pattern of clustering among private rooms on Airbnb in Singapore.

Clark and evans test for Shared rooms

The test hypotheses are:

Null Hypothesis: The distribution of shared rooms on Airbnb in Singapore are randomly distributed.

Alternate Hypothesis: The distribution of shared rooms on Airbnb in Singapore are not randomly distributed.

The 95% confidence interval will be used.

clarkevans.test(shared_ppp_jit,
                correction="none",
                clipregion="sg_owin",
                alternative=c("two.sided"),
                nsim=99)

## 
##  Clark-Evans test
##  No edge correction
##  Monte Carlo test based on 99 simulations of CSR with fixed n
## 
## data:  shared_ppp_jit
## R = 0.34303, p-value = 0.02
## alternative hypothesis: two-sided

The p-value (0.002) is less than 0.05, hence we can reject the null hypothesis and conclude that we are 95% confident that the distribution of shared rooms on Airbnb in Singapore are not randomly distributed. Furthermore, as the R value (0.40) is less than one, there is a significant pattern of clustering among shared rooms on Airbnb in Singapore.

Conclusion

All the four room types show significant patterns of clustering, with hotel rooms exhibiting the most amount of clustering and shared rooms exhibiting the least amount of clustering. To gain a deeper insight on the spatial arrangement of the different room types, the data will be analysed through kernel density maps.

Kernel Density Estimation

Plotting kernal density maps

par(mfrow = c(2, 2))

home_ppp_jit.km <- rescale(home_ppp_jit, 1000, "km")
hotel_ppp_jit.km <- rescale(hotel_ppp_jit, 1000, "km")
shared_ppp_jit.km <- rescale(shared_ppp_jit, 1000, "km")
private_ppp_jit.km <- rescale(private_ppp_jit, 1000, "km")

kde_home_bw <- density(home_ppp_jit.km, sigma=bw.scott, edge=FALSE, kernel="gaussian")
plot(kde_home_bw, main="Full home/apartment kernel density")
kde_hotel_bw <- density(hotel_ppp_jit.km, sigma=bw.scott, edge=FALSE, kernel="gaussian")
plot(kde_hotel_bw, main="Hotel rooms kernel density")
kde_shared_bw <- density(shared_ppp_jit.km, sigma=bw.scott, edge=FALSE, kernel="gaussian")
plot(kde_shared_bw, main="Shared rooms kernel density")
kde_private_bw <- density(private_ppp_jit.km, sigma=bw.scott, edge=FALSE, kernel="gaussian")
plot(kde_private_bw, main="Private rooms kernel density")

kde_home_unscalled <- density(home_ppp_jit, sigma=bw.scott, edge=FALSE, kernel="gaussian")
kde_hotel_unscalled <- density(hotel_ppp_jit, sigma=bw.scott, edge=FALSE, kernel="gaussian")
kde_shared_unscalled <- density(shared_ppp_jit, sigma=bw.scott, edge=FALSE, kernel="gaussian")
kde_private_unscalled <- density(private_ppp_jit, sigma=bw.scott, edge=FALSE, kernel="gaussian")

From the above kernel density maps, we can derive that all the room types are concentrated in the central region of Singapore. However, colour in each of the maps are in reference to its own scale and are not for comparison purposes. The KDEs are plotted against each other to identify the region in which each of these room types are concentrated in.

To understand these densities better, the KDE is converted into a raster layer and then plotted on an openstreetmap in order to spatially understand the data better.

Conversion of gridded objectss into raster data

kde_home_raster <- raster(kde_home_unscalled)
kde_private_raster <- raster(kde_private_unscalled)
kde_shared_raster <- raster(kde_shared_unscalled)
kde_hotel_raster <- raster(kde_hotel_unscalled)

projection(kde_home_raster) <- CRS("+init=EPSG:3414")
projection(kde_private_raster) <- CRS("+init=EPSG:3414")
projection(kde_shared_raster) <- CRS("+init=EPSG:3414")
projection(kde_hotel_raster) <- CRS("+init=EPSG:3414")

values(kde_home_raster) <- values(kde_home_raster)*1000000
values(kde_private_raster) <- values(kde_private_raster)*1000000
values(kde_shared_raster) <- values(kde_shared_raster)*1000000
values(kde_hotel_raster) <- values(kde_hotel_raster)*1000000

Viewing airbnb listing by room type in Singapore

tmap_mode("view")
a<-tm_shape(kde_home_raster)+
  tm_raster(colorNA = "lightgrey", showNA = FALSE,alpha=0.5,title = "Number of listings")+
  tm_basemap(group = "OpenStreetMap")+
  tm_view(set.zoom.limits = c(10,18)) 

b<-tm_shape(kde_private_raster)+
  tm_raster(colorNA = "lightgrey", showNA = FALSE,alpha=0.5,title = "Number of listings")+
    tm_basemap(group = "OpenStreetMap")+
    tm_view(set.zoom.limits = c(10,18)) 


c<-tm_shape(kde_shared_raster)+
  tm_raster(colorNA = "lightgrey", showNA = FALSE,alpha=0.5,title = "Number of listings")+
    tm_basemap(group = "OpenStreetMap")+
    tm_view(set.zoom.limits = c(10,18)) 


d<-tm_shape(kde_hotel_raster)+
  tm_raster(colorNA = "lightgrey", showNA = FALSE,alpha=0.5,title = "Number of listings")+
    tm_basemap(group = "OpenStreetMap")+
    tm_view(set.zoom.limits = c(10,18)) 


tmap_arrange(a,b,c,d)
tmap_mode("plot")

!(/Users/Amey/Desktop/Y2S2/IS415/assign2/data/aspatial/maps/m1.png)

Interpretation and Significance of Kernel Density Maps

From the above map, it is evident that the central subzones have the most proportion of each of the room types as compared to the other regions of Singapore. Using Kernel density map over the point map has allowed to us to gain more insights. As there are cummulatively thousands of listings for each room type, the Kernel density map allows one point to describe multiple observations in the nearby proximity. When looking at Singapore as a whole, this technique allows us to understand the spread of the data properly. Furthermore, the point data does not allow to gain relative understanding of the distribution of the data as the map gets very cluttered when the number of points are in thousands.

Section B: Analysis by planning subzones

Exploratory Data Analysis

Data preperation by extracting subzones

mpsz <- st_set_crs(mpsz,3414)
mpsz2 <- st_transform(mpsz,3414)
aljunied <- mpsz2 %>% filter(SUBZONE_N=="ALJUNIED")
balestier <- mpsz2 %>% filter(SUBZONE_N=="BALESTIER")
lavender <- mpsz2 %>% filter(SUBZONE_N=="LAVENDER")
tanjongpagar <-  mpsz2 %>% filter(SUBZONE_N=="TANJONG PAGAR")

## Area
a1 <- st_area(aljunied)/1000000
a2 <- st_area(balestier)/1000000
a3 <- st_area(lavender)/1000000
a4 <- st_area(tanjongpagar)/1000000

Visualising Airbnb listing by room type within Aljunied

tmap_mode("plot")

## tmap mode set to plotting

data <- st_intersection(airbnb3414,aljunied)
data_aljunied <- st_jitter(data)

tm_shape(aljunied)+
  tm_borders()+
  tm_shape(data_aljunied)+
  tm_facets(by="room_type",free.coords = FALSE)+
  tm_dots("room_type",size=0.4,alpha=0.7,title = "Room Type")+
  tm_layout(main.title="Distribution of Airbnb listings by room type in Aljunied")

We will now examine the spatial patterns observed in each of the distributions. Entire home/apartment are concentrated in the middle of Aljunied, and has fewer options in the northern and southern region of Aljunied. Private rooms are located in nearly all the parts of Aljunied, however, has a fewer number of options in the northern region. The availibility of hotel rooms and shared rooms are much lower as compared to the previous two categories. Hotel rooms are found in the central region of Aljunied only whereas shared rooms are scattered over the central and eastern region of Aljunied.

To gain statistical understanding of the distribution, K function is performed on every room type to determine if there is complete spatial randomness.

Data preperation for each room type

aljunied_home <- data_aljunied%>%filter(room_type=="Entire home/apt")
a<- as(aljunied_home,"Spatial")
b <- as(a,"SpatialPoints")
aljunied_home_ppp <- as(b,"ppp")

aljunied_proom <- data_aljunied%>%filter(room_type=="Private room")
a<- as(aljunied_proom,"Spatial")
b <- as(a,"SpatialPoints")
aljunied_proom_ppp <- as(b,"ppp")

aljunied_sroom <- data_aljunied%>%filter(room_type=="Shared room")
a<- as(aljunied_sroom,"Spatial")
b <- as(a,"SpatialPoints")
aljunied_sroom_ppp <- as(b,"ppp")

aljunied_hotel <- data_aljunied%>%filter(room_type=="Hotel room")
a<- as(aljunied_hotel,"Spatial")
b <- as(a,"SpatialPoints")
aljunied_hotel_ppp <- as(b,"ppp")

Second order spatial analysis in Aljunied through K function

k_aljunied_home = Kest(aljunied_home_ppp, correction = "Ripley")
plot(k_aljunied_home, .-r ~ r, ylab= "K(d)-r", xlab = "d(m)", main="K-function plot of Entire home/apt in Aljunied")

As seen in the above graph, as the K-function value is over its expected value, it shows signs of clustering. However, to ensure that the clustering is statistically significant, we will perform a complete spatial randomness test.

CSR test for Full home/apt in Aljunied

The test hypothesis is as follows:

Null hypothesis: The distribution of Entire home/apt listings in Aljunied are randomly distributed.

Alternate hypothesis: The distribution of Entire home/apt listings in Aljunied are not randomly distributed.

The null hypothesis will be rejected if p-value is smaller than alpha value of 0.001.

k_aljunied_home.csr <- envelope(aljunied_home_ppp, Kest, nsim = 999, nrank = 1, glocal = TRUE)

## Generating 999 simulations of CSR  ...
## 1, 2, 3, ......10.........20.........30.........40.........50.........60........
## .70.........80.........90.........100.........110.........120.........130......
## ...140.........150.........160.........170.........180.........190.........200....
## .....210.........220.........230.........240.........250.........260.........270..
## .......280.........290.........300.........310.........320.........330.........340
## .........350.........360.........370.........380.........390.........400........
## .410.........420.........430.........440.........450.........460.........470......
## ...480.........490.........500.........510.........520.........530.........540....
## .....550.........560.........570.........580.........590.........600.........610..
## .......620.........630.........640.........650.........660.........670.........680
## .........690.........700.........710.........720.........730.........740........
## .750.........760.........770.........780.........790.........800.........810......
## ...820.........830.........840.........850.........860.........870.........880....
## .....890.........900.........910.........920.........930.........940.........950..
## .......960.........970.........980.........990........ 999.
## 
## Done.

plot(k_aljunied_home.csr, . - r ~ r, xlab="d", ylab="K(d)-r",  main="CSR test on Full home/apt in Aljunied")

As the observed k-value curve lies outside of the envelope, the null hypothesis is rejected and we can conclude that the distribution of Entire home/apt listings in Aljunied are not randomly distributed with 99.9% confidence. Furthermore, as the observed K function curve is above the envelope, the distribution of “Entire home/apt” listings on Airbnb in Aljunied subzone exhibit clustering. As the curve is not extremely steep, there are no significant cluster locations in Aljunied, indicating that most of the properties are in close proximity to one another as seen in the map as well.

Distribution of Private Room in Aljunied

k_aljunied_proom = Kest(aljunied_proom_ppp, correction = "Ripley")
plot(k_aljunied_proom, .-r ~ r, ylab= "K(d)-r", xlab = "d(m)", main="K-function plot of Private Room in Aljunied")

As seen in the above graph, as the K-function value is above its expected value, it shows signs of clustering. However, to ensure that the clustering is statistically significant, we will perform a complete spatial randomness test.

CSR test on Private Rooms in Aljunied

The test hypothesis is as follows:

Null hypothesis: The distribution of Private room listings in Aljunied are randomly distributed.

Alternate hypothesis: The distribution of Private room listings in Aljunied are not randomly distributed.

The null hypothesis will be rejected if p-value is smaller than alpha value of 0.001.

k_aljunied_proom.csr <- envelope(aljunied_proom_ppp, Kest, nsim = 999, nrank = 1, glocal=TRUE)

## Generating 999 simulations of CSR  ...
## 1, 2, 3, ......10.........20.........30.........40.........50.........60........
## .70.........80.........90.........100.........110.........120.........130......
## ...140.........150.........160.........170.........180.........190.........200....
## .....210.........220.........230.........240.........250.........260.........270..
## .......280.........290.........300.........310.........320.........330.........340
## .........350.........360.........370.........380.........390.........400........
## .410.........420.........430.........440.........450.........460.........470......
## ...480.........490.........500.........510.........520.........530.........540....
## .....550.........560.........570.........580.........590.........600.........610..
## .......620.........630.........640.........650.........660.........670.........680
## .........690.........700.........710.........720.........730.........740........
## .750.........760.........770.........780.........790.........800.........810......
## ...820.........830.........840.........850.........860.........870.........880....
## .....890.........900.........910.........920.........930.........940.........950..
## .......960.........970.........980.........990........ 999.
## 
## Done.

plot(k_aljunied_proom.csr, . - r ~ r, xlab="d", ylab="K(d)-r",  main="CSR test on Private rooms in Aljunied")

As the observed k-value curve lies above the envelope, the null hypothesis is rejected and we can conclude that the distribution of Private room listings in Aljunied are not randomly distributed with 99.9% confidence. Furthermore, as the observed K function curve is above the envelope, the distribution of “Private Room” listings on Airbnb in Aljunied subzone significantly exhibit clustering. As the curve is not extremely steep, there are no significant cluster locations in Aljunied, indicating that most of the properties are in close proximity to one another as seen in the map as well.

Distribution of Shared Room in Aljunied

k_aljunied_sroom = Kest(aljunied_sroom_ppp, correction = "Ripley")
plot(k_aljunied_sroom, .-r ~ r, ylab= "K(d)-r", xlab = "d(m)", main="K-function plot of Shared Room in Aljunied")

As seen in the above graph, as the K-function value is over its expected value for most of the d(m) values except between m=80 to m=120, hence shows signs of clustering. However, to ensure that the clustering is statistically significant, we will perform a complete spatial randomness test.

The test hypothesis is as follows:

Null hypothesis: The distribution of Shared room listings in Aljunied are randomly distributed.

Alternate hypothesis: The distribution of Shared room listings in Aljunied are not randomly distributed.

The null hypothesis will be rejected if p-value is smaller than alpha value of 0.001.

k_aljunied_sroom.csr <- envelope(aljunied_sroom_ppp, Kest, nsim = 999, nrank = 1, glocal=TRUE)

## Generating 999 simulations of CSR  ...
## 1, 2, 3, ......10.........20.........30.........40.........50.........60........
## .70.........80.........90.........100.........110.........120.........130......
## ...140.........150.........160.........170.........180.........190.........200....
## .....210.........220.........230.........240.........250.........260.........270..
## .......280.........290.........300.........310.........320.........330.........340
## .........350.........360.........370.........380.........390.........400........
## .410.........420.........430.........440.........450.........460.........470......
## ...480.........490.........500.........510.........520.........530.........540....
## .....550.........560.........570.........580.........590.........600.........610..
## .......620.........630.........640.........650.........660.........670.........680
## .........690.........700.........710.........720.........730.........740........
## .750.........760.........770.........780.........790.........800.........810......
## ...820.........830.........840.........850.........860.........870.........880....
## .....890.........900.........910.........920.........930.........940.........950..
## .......960.........970.........980.........990........ 999.
## 
## Done.

plot(k_aljunied_sroom.csr, . - r ~ r, xlab="d", ylab="K(d)-r",  main="CSR test on Shared rooms in Aljunied")

As the observed k-value curve lies in the envelope region, the null hypothesis is not rejected and we can conclude that the distribution of Shared room listings in Aljunied are randomly distributed with 99.9% confidence. However, as the number of Shared rooms in Aljunied are only 10, the K function is extremely sensitive to each of the points.

Distribution of Hotel Room in Aljunied

k_aljunied_hotel = Kest(aljunied_hotel_ppp, correction = "Ripley")
plot(k_aljunied_hotel, .-r ~ r, ylab= "K(d)-r", xlab = "d(m)", main="K-function plot of Hotel Room in Aljunied")

As seen in the above graph, as the K-function value is over its expected value, it shows signs of clustering. As the k-function is increasing in “steps” from d=15, there are signs of equal distribution of hotel rooms in the Aljunied area. However, to ensure that the clustering is statistically significant, we will perform a complete spatial randomness test. The test hypothesis is as follows:

Null hypothesis: The distribution of Hotel room listings in Aljunied are randomly distributed.

Alternate hypothesis: The distribution of Hotel room listings in Aljunied are not randomly distributed.

The null hypothesis will be rejected if p-value is smaller than alpha value of 0.001.

k_aljunied_hotel.csr <- envelope(aljunied_hotel_ppp, Kest, nsim = 999, nrank = 1,glocal = TRUE)

## Generating 999 simulations of CSR  ...
## 1, 2, 3, ......10.........20.........30.........40.........50.........60........
## .70.........80.........90.........100.........110.........120.........130......
## ...140.........150.........160.........170.........180.........190.........200....
## .....210.........220.........230.........240.........250.........260.........270..
## .......280.........290.........300.........310.........320.........330.........340
## .........350.........360.........370.........380.........390.........400........
## .410.........420.........430.........440.........450.........460.........470......
## ...480.........490.........500.........510.........520.........530.........540....
## .....550.........560.........570.........580.........590.........600.........610..
## .......620.........630.........640.........650.........660.........670.........680
## .........690.........700.........710.........720.........730.........740........
## .750.........760.........770.........780.........790.........800.........810......
## ...820.........830.........840.........850.........860.........870.........880....
## .....890.........900.........910.........920.........930.........940.........950..
## .......960.........970.........980.........990........ 999.
## 
## Done.

plot(k_aljunied_hotel.csr, . - r ~ r, xlab="d", ylab="K(d)-r",  main="CSR test on Hotel rooms in Aljunied")

As the observed most part of the k-value curve lies in the envelope, the null hypothesis is not rejected and we can conclude that the distribution of Hotel room listings in Aljunied are randomly distributed with 99.9% confidence. The k-function is outside the envelope for two short intervals between d=30 and d=601, exhibiting few significant signs of clustering.

Insights from CSR tests in Aljunied

From the CSR tests conducted above, the room type “Full Apt/Home” and “Private Room” exhibit significant clustering whereas “Hotel room” and “Shared room” exhibit complete spatial randomness. However, a point to consider is that the number of shared rooms and hotel rooms are much lesser as compared to the number of full home/apt and private room listings which maybe one of the external factors contributing to the conclusions made as the K-function is very sensitive to the number of observations.

To analyse these patterns further, kernel density maps are drawn for each Room type.

Kernel density maps by room type in Aljunied

a <- as(aljunied,"Spatial")
aljunied_owin <- as(a,"owin")
par(mfrow = c(2, 2))


# Full home/apt
aljunied_home_ppp <- aljunied_home_ppp[aljunied_owin]
aljunied_home_ppp.km <- rescale(aljunied_home_ppp, 1000, "km")

kde_aljunied_home <- density(aljunied_home_ppp.km, sigma=0.25, edge=TRUE, kernel="gaussian")
plot(kde_aljunied_home)
kde_aljunied_home_unscalled <- density(aljunied_home_ppp, sigma=bw.scott, edge=TRUE, kernel="gaussian")

# Private room
aljunied_proom_ppp <- aljunied_proom_ppp[aljunied_owin]
aljunied_proom_ppp.km <- rescale(aljunied_proom_ppp, 1000, "km")

kde_aljunied_proom <- density(aljunied_proom_ppp.km, sigma=0.25, edge=TRUE, kernel="gaussian")
plot(kde_aljunied_proom)
kde_aljunied_proom_unscalled <- density(aljunied_proom_ppp, sigma=bw.scott, edge=TRUE, kernel="gaussian")

# Shared room
aljunied_sroom_ppp <- aljunied_sroom_ppp[aljunied_owin]
aljunied_sroom_ppp.km <- rescale(aljunied_sroom_ppp, 1000, "km")

kde_aljunied_sroom <- density(aljunied_sroom_ppp.km, sigma=0.25, edge=TRUE, kernel="gaussian")
plot(kde_aljunied_sroom)
kde_aljunied_sroom_unscalled <- density(aljunied_sroom_ppp, sigma=bw.scott, edge=TRUE, kernel="gaussian")

# Hotel room
aljunied_hotel_ppp <- aljunied_hotel_ppp[aljunied_owin]
aljunied_hotel_ppp.km <- rescale(aljunied_hotel_ppp, 1000, "km")

kde_aljunied_hotel <- density(aljunied_hotel_ppp.km, sigma=0.25, edge=TRUE, kernel="gaussian")
plot(kde_aljunied_hotel)

kde_aljunied_hotel_unscalled <- density(aljunied_hotel_ppp, sigma=bw.scott, edge=TRUE, kernel="gaussian")

Data conversion to raster and assigning it projection

kde_aljunied_home_raster <- raster(kde_aljunied_home_unscalled)
projection(kde_aljunied_home_raster) <- CRS("+init=EPSG:3414")
values(kde_aljunied_home_raster) <- values(kde_aljunied_home_raster)*100000

kde_aljunied_proom_raster <- raster(kde_aljunied_proom_unscalled)
projection(kde_aljunied_proom_raster) <- CRS("+init=EPSG:3414")
values(kde_aljunied_proom_raster) <- values(kde_aljunied_proom_raster)*100000

kde_aljunied_sroom_raster <- raster(kde_aljunied_sroom_unscalled)
projection(kde_aljunied_sroom_raster) <- CRS("+init=EPSG:3414")
values(kde_aljunied_sroom_raster) <- values(kde_aljunied_sroom_raster)*1000000

kde_aljunied_hotel_raster <- raster(kde_aljunied_hotel_unscalled)
projection(kde_aljunied_hotel_raster) <- CRS("+init=EPSG:3414")
values(kde_aljunied_hotel_raster) <- values(kde_aljunied_hotel_raster)*100000

Visualising the raster data on openstreemap

The code below produces raster layer data on Openstreetmap. However, to do so, the interactive view was used. For publishing purposes on RPubs, screenshots of the maps are taken and attached after the code.

# Full house/apt
tmap_mode("view")
tm_shape(kde_aljunied_home_raster)+
  tm_raster(colorNA = "lightgrey", showNA = FALSE,alpha=0.5,title = "Number of listings")+
  tm_shape(aljunied)+
  tm_borders(lwd = 2)+
  tm_basemap(group = "OpenStreetMap")+
  tm_layout("Distribution of Full home/apt in Aljunied")

# Private room
tm_shape(kde_aljunied_proom_raster)+
  tm_raster(colorNA = "lightgrey", showNA = FALSE,alpha=0.5,title = "Number of listings")+
  tm_shape(aljunied)+
  tm_borders(lwd = 2)+
  tm_basemap(group = "OpenStreetMap")+
  tm_layout("Distribution of Private room in Aljunied")

# Shared room
tm_shape(kde_aljunied_sroom_raster)+
  tm_raster(colorNA = "lightgrey", showNA = FALSE,alpha=0.5,title = "Number of listings")+
  tm_shape(aljunied)+
  tm_borders(lwd = 2)+
  tm_basemap(group = "OpenStreetMap")+
  tm_layout("Distribution of Shared room in Aljunied")

# Hotel room
tm_shape(kde_aljunied_hotel_raster)+
  tm_raster(colorNA = "lightgrey", showNA = FALSE,alpha=0.5,title = "Number of listings", palette = "YlOrBr")+
  tm_shape(aljunied)+
  tm_borders(lwd = 2)+
  tm_basemap(group = "OpenStreetMap")+
  tm_layout("Distribution of Hotel room in Aljunied")

Screenshots of maps

Interpretation

There are two significant groups of locations for Full apt/home listings which are in close proximity of Aljunied MRT. The northern region of Aljunied does not have many full apt/home listings because it mainly comprises of HDBs, where leasing out via Airbnb is illegal. There are various shared rooms present in close proximity of Aljunied MRT and Paya Lebar MRT as compared to rest of the region which is not so well connected with the MRT network. Most of the hotels present in the Aljunied subzone are on the same road (Geylang Road) as there are various small chain of hotels and lodges present in this area as compared to rest of the region, which is predominantly residential.

Visualising Airbnb listing by room type within Balestier

data <- st_intersection(airbnb3414,balestier)
data_balestier <- st_jitter(data)

tmap_mode("plot")

## tmap mode set to plotting

tm_shape(balestier)+
  tm_borders()+
  tm_shape(data)+
  tm_facets(by="room_type",free.coords = FALSE)+
  tm_dots("room_type",size=0.5,alpha=0.7,title = "Room Type")+
  tm_layout(main.title="Distribution of Airbnb listings by room type in Balestier")

We will now examine the spatial patterns observed in each of the distributions. Entire home/apartment listings are present in all regions except the eastern part of Balestier. Private rooms and Hotel rooms are located in nearly all the parts of Balestier, however, has a fewer number of options in the eastern region. There is only one shared room in balestier, hence we won’t be performing any spatial analysis on it.

To gain statistical understanding of the distribution, K function is performed on every room type except shared room to determine if there is complete spatial randomness.

Data preperation for every room type

balestier_home <- data_balestier%>%filter(room_type=="Entire home/apt")
a<- as(balestier_home,"Spatial")
b <- as(a,"SpatialPoints")
balestier_home_ppp <- as(b,"ppp")

balestier_hotel <- data_balestier%>%filter(room_type=="Hotel room")
a<- as(balestier_hotel,"Spatial")
b <- as(a,"SpatialPoints")
balestier_hotel_ppp <- as(b,"ppp")

balestier_proom <- data_balestier%>%filter(room_type=="Private room")
a<- as(balestier_proom,"Spatial")
b <- as(a,"SpatialPoints")
balestier_proom_ppp <- as(b,"ppp")

Second order spatial analysis in Balestier through K function

Distribution of Full home/apt in Balestier through K function

k_balestier_home = Kest(aljunied_home_ppp, correction = "Ripley")
plot(k_balestier_home, .-r ~ r, ylab= "K(d)-r", xlab = "d(m)", main="K-function plot of Entire home/apt in Balestier")

Performing CSR test

The test hypothesis is as follows:

Null hypothesis: The distribution of Entire home/apt listings in Balestier are randomly distributed.

Alternate hypothesis: The distribution of Entire home/apt listings in Balestier are not randomly distributed.

The null hypothesis will be rejected if p-value is smaller than alpha value of 0.001.

k_balestier_home.csr <- envelope(balestier_home_ppp, Kest, nsim = 999, nrank = 1,glocal=TRUE)

## Generating 999 simulations of CSR  ...
## 1, 2, 3, ......10.........20.........30.........40.........50.........60........
## .70.........80.........90.........100.........110.........120.........130......
## ...140.........150.........160.........170.........180.........190.........200....
## .....210.........220.........230.........240.........250.........260.........270..
## .......280.........290.........300.........310.........320.........330.........340
## .........350.........360.........370.........380.........390.........400........
## .410.........420.........430.........440.........450.........460.........470......
## ...480.........490.........500.........510.........520.........530.........540....
## .....550.........560.........570.........580.........590.........600.........610..
## .......620.........630.........640.........650.........660.........670.........680
## .........690.........700.........710.........720.........730.........740........
## .750.........760.........770.........780.........790.........800.........810......
## ...820.........830.........840.........850.........860.........870.........880....
## .....890.........900.........910.........920.........930.........940.........950..
## .......960.........970.........980.........990........ 999.
## 
## Done.

plot(k_balestier_home.csr, . - r ~ r, xlab="d", ylab="K(d)-r", main="CSR test on full apt/home in Balestier")

As the observed k-value curve lies above the envelope, the null hypothesis is rejected and we can conclude that the distribution of Entire home/apt listings in Balestier are not randomly distributed with 99.9% confidence. Furthermore, as the observed K function curve is above the envelope, the distribution of “Entire home/apt” listings on Airbnb in Balestier subzone exhibit clustering. As the curve is not extremely steep, there are no significant cluster locations in Balestier, indicating that most of the properties are in close proximity to one another as seen in the map as well.

Distribution of Hotel room in Balestier

k_balestier_hotel = Kest(balestier_hotel_ppp, correction = "Ripley", glocal=TRUE)
plot(k_balestier_hotel, .-r ~ r, ylab= "K(d)-r", xlab = "d(m)", main="K-function plot of Hotel Room in Balestier")

CSR test

The test hypothesis is as follows:

Null hypothesis: The distribution of Hotel room listings in Balestier are randomly distributed.

Alternate hypothesis: The distribution of Hotel room listings in Balestier are not randomly distributed.

The null hypothesis will be rejected if p-value is smaller than alpha value of 0.001.

k_balestier_hotel.csr <- envelope(balestier_hotel_ppp, Kest, nsim = 999, nrank = 1)

## Generating 999 simulations of CSR  ...
## 1, 2, 3, ......10.........20.........30.........40.........50.........60........
## .70.........80.........90.........100.........110.........120.........130......
## ...140.........150.........160.........170.........180.........190.........200....
## .....210.........220.........230.........240.........250.........260.........270..
## .......280.........290.........300.........310.........320.........330.........340
## .........350.........360.........370.........380.........390.........400........
## .410.........420.........430.........440.........450.........460.........470......
## ...480.........490.........500.........510.........520.........530.........540....
## .....550.........560.........570.........580.........590.........600.........610..
## .......620.........630.........640.........650.........660.........670.........680
## .........690.........700.........710.........720.........730.........740........
## .750.........760.........770.........780.........790.........800.........810......
## ...820.........830.........840.........850.........860.........870.........880....
## .....890.........900.........910.........920.........930.........940.........950..
## .......960.........970.........980.........990........ 999.
## 
## Done.

plot(k_balestier_hotel.csr, . - r ~ r, xlab="d", ylab="K(d)-r", main="CSR test on Hotel room in Balestier")

The observed k-function curve lies in the envelope from d=0 to d=120(with a small portion inside the envelope in between), after which the curve lies out of the envelope. Therefore, we can reject the null hypothesis for d>120, and conclude that the distribution of Hotel room listings in Balestier are not randomly distributed for d>120 with 99.9% confidence.

Distribution of Private room in Balestier

k_balestier_proom = Kest(balestier_proom_ppp, correction = "Ripley", glocal=TRUE)
plot(k_balestier_proom, .-r ~ r, ylab= "K(d)-r", xlab = "d(m)", main="K-function plot of Private room in balestier")

CSR test

The test hypothesis is as follows:

Null hypothesis: The distribution of Private room listings in Balestier are randomly distributed.

Alternate hypothesis: The distribution of Private room listings in Balestier are not randomly distributed.

The null hypothesis will be rejected if p-value is smaller than alpha value of 0.001.

k_balestier_proom.csr <- envelope(balestier_proom_ppp, Kest, nsim = 999, nrank = 1)

## Generating 999 simulations of CSR  ...
## 1, 2, 3, ......10.........20.........30.........40.........50.........60........
## .70.........80.........90.........100.........110.........120.........130......
## ...140.........150.........160.........170.........180.........190.........200....
## .....210.........220.........230.........240.........250.........260.........270..
## .......280.........290.........300.........310.........320.........330.........340
## .........350.........360.........370.........380.........390.........400........
## .410.........420.........430.........440.........450.........460.........470......
## ...480.........490.........500.........510.........520.........530.........540....
## .....550.........560.........570.........580.........590.........600.........610..
## .......620.........630.........640.........650.........660.........670.........680
## .........690.........700.........710.........720.........730.........740........
## .750.........760.........770.........780.........790.........800.........810......
## ...820.........830.........840.........850.........860.........870.........880....
## .....890.........900.........910.........920.........930.........940.........950..
## .......960.........970.........980.........990........ 999.
## 
## Done.

plot(k_balestier_proom.csr, . - r ~ r, xlab="d", ylab="K(d)-r", main="CSR test on Private room in Balestier")

As the observed k-value curve lies above the envelope for all values of d, the null hypothesis is rejected and we can conclude that the distribution of Private room listings in Balestier are not randomly distributed with 99.9% confidence. Furthermore, as the observed K function curve is above the envelope, the distribution of Private rooms on Airbnb in Balestier subzone significantly exhibit clustering.

To analyse these patterns further, kernel density maps are drawn for each Room type.

Data preperation

a <- as(balestier,"Spatial")
balestier_owin <- as(a,"owin")

balestier_home_ppp <- balestier_home_ppp[balestier_owin]
balestier_home_ppp.km <- rescale(balestier_home_ppp, 1000, "km")

balestier_hotel_ppp <- balestier_hotel_ppp[balestier_owin]
balestier_hotel_ppp.km <- rescale(balestier_hotel_ppp, 1000, "km")

balestier_proom_ppp <- balestier_proom_ppp[balestier_owin]
balestier_proom_ppp.km <- rescale(balestier_proom_ppp, 1000, "km")

Plotting Kernel density

par(mfrow = c(2, 2))

kde_balestier_home <- density(balestier_home_ppp.km, sigma=0.25, edge=TRUE, kernel="gaussian")
plot(kde_balestier_home)
kde_balestier_home_unscalled <- density(balestier_home_ppp, sigma=bw.scott, edge=TRUE, kernel="gaussian")

kde_balestier_hotel <- density.ppp(balestier_hotel_ppp.km, sigma=0.25, edge=TRUE, kernel="gaussian")
plot(kde_balestier_hotel)
kde_balestier_hotel_unscalled <- density(balestier_hotel_ppp, sigma=bw.scott, edge=TRUE, kernel="gaussian")

kde_balestier_proom <- density(balestier_proom_ppp.km, sigma=0.25, edge=TRUE, kernel="gaussian")
plot(kde_balestier_proom)
kde_balestier_proom_unscalled <- density(balestier_proom_ppp, sigma=bw.scott, edge=TRUE, kernel="gaussian")

As seen in all the Kernel density maps above, a clear pattern can be observed. All the three room types are concentrated in the central region of Balestier, with very few listings on the Eastern and Western parts of Balestier.

To determine the cause of this divide, the kernel density map will be converted into a raster layer and then plotted into a base layer of OpenStreetMap to see the how various spatial elements in Balestier affect where the listings are present.

Data conversion to raster and assigning it projection

kde_balestier_home_raster <- raster(kde_balestier_home_unscalled)
projection(kde_balestier_home_raster) <- CRS("+init=EPSG:3414")
values(kde_balestier_home_raster) <- values(kde_balestier_home_raster)*100000

kde_balestier_hotel_raster <- raster(kde_balestier_hotel_unscalled)
projection(kde_balestier_hotel_raster) <- CRS("+init=EPSG:3414")
values(kde_balestier_hotel_raster) <- values(kde_balestier_hotel_raster)*100000

kde_balestier_proom_raster <- raster(kde_balestier_proom_unscalled)
projection(kde_balestier_proom_raster) <- CRS("+init=EPSG:3414")
values(kde_balestier_proom_raster) <- values(kde_balestier_proom_raster)*100000

Visualising the raster data on openstreemap

tmap_mode("view")
tm_shape(kde_balestier_home_raster)+
  tm_raster(colorNA = "lightgrey", showNA = FALSE,alpha=0.5,title = "Number of listings")+
  tm_shape(balestier)+
  tm_borders(lwd = 2)+
  tm_basemap(group = "OpenStreetMap")+
  tm_layout("Distribution of Full home/apt in Balestier")

tm_shape(kde_balestier_hotel_raster)+
  tm_raster(colorNA = "lightgrey", showNA = FALSE,alpha=0.5,title = "Number of listings")+
  tm_shape(balestier)+
  tm_borders(lwd = 2)+
  tm_basemap(group = "OpenStreetMap")+
  tm_layout("Distribution of Hotel room in Balestier")

tm_shape(kde_balestier_proom_raster)+
  tm_raster(colorNA = "lightgrey", showNA = FALSE,alpha=0.5,title = "Number of listings")+
  tm_shape(balestier)+
  tm_borders(lwd = 2)+
  tm_basemap(group = "OpenStreetMap")+
  tm_layout("Distribution of Private room in Balestier")
tmap_mode("plot")

Screenshots of maps

Interpretation

There are two significant groups of locations for Full apt/home listings which are both located on Balestier Road. This is due to the fact that this subzone falls very close to the city, and hence consists of various other essential services such as hospitals, offices and shopping centres. The shared rooms are present in the eastern region of the zone as many residential facilities are present here. The hotel rooms are present in the same area as the full hom/apartment and shared rooms, as the other regions mainly consist of hospitals, offices and plazas.

Visualising Airbnb listing by room type within Lavender

tmap_mode("plot")

## tmap mode set to plotting

data <- st_intersection(airbnb3414,lavender)
data_lavender <- st_jitter(data)

tm_shape(lavender)+
  tm_borders()+
  tm_shape(data_lavender)+
  tm_facets(by="room_type",free.coords = FALSE)+
  tm_dots("room_type",size=0.5,alpha=0.7, title = "Room type")+
  tm_layout(main.title="Distribution of Airbnb listings by room type in Lavender")

We will now examine the spatial patterns observed in each of the distributions. Entire home/apartment are concentrated in the northern region of Lavender, and has fewer options in the southern region of Lavender. Private rooms are located in nearly all the parts of Lavender. The availibility of hotel rooms and shared rooms are much lower as compared to the previous two categories. Both are found in the central region and northern region of Lavender.

To gain statistical understanding of the distribution, K function is performed on every room type to determine if there is complete spatial randomness.

Data preperation for each room type

lavender_home <- data_lavender%>%filter(room_type=="Entire home/apt")
a<- as(lavender_home,"Spatial")
b <- as(a,"SpatialPoints")
lavender_home_ppp <- as(b,"ppp")

lavender_proom <- data_lavender%>%filter(room_type=="Private room")
a<- as(lavender_proom,"Spatial")
b <- as(a,"SpatialPoints")
lavender_proom_ppp <- as(b,"ppp")

lavender_sroom <- data_lavender%>%filter(room_type=="Shared room")
a<- as(lavender_sroom,"Spatial")
b <- as(a,"SpatialPoints")
lavender_sroom_ppp <- as(b,"ppp")

lavender_hotel <- data_lavender%>%filter(room_type=="Hotel room")
a<- as(lavender_hotel,"Spatial")
b <- as(a,"SpatialPoints")
lavender_hotel_ppp <- as(b,"ppp")

Second order spatial analysis in Lavender through K function

Distribution of Entire home/apt in Lavender

k_lavender_home = Kest(lavender_home_ppp, correction = "Ripley")
plot(k_lavender_home, .-r ~ r, ylab= "K(d)-r", xlab = "d(m)", main="K-function plot of Entire home/apt in lavender")

CSR test

The test hypothesis is as follows:

Null hypothesis: The distribution of Entire home/apt listings in Lavender are randomly distributed.

Alternate hypothesis: The distribution of Entire home/apt listings in Lavender are not randomly distributed.

The null hypothesis will be rejected if p-value is smaller than alpha value of 0.001.

k_lavender_home.csr <- envelope(lavender_home_ppp, Kest, nsim = 999, nrank = 1,glocal=TRUE)

## Generating 999 simulations of CSR  ...
## 1, 2, 3, ......10.........20.........30.........40.........50.........60........
## .70.........80.........90.........100.........110.........120.........130......
## ...140.........150.........160.........170.........180.........190.........200....
## .....210.........220.........230.........240.........250.........260.........270..
## .......280.........290.........300.........310.........320.........330.........340
## .........350.........360.........370.........380.........390.........400........
## .410.........420.........430.........440.........450.........460.........470......
## ...480.........490.........500.........510.........520.........530.........540....
## .....550.........560.........570.........580.........590.........600.........610..
## .......620.........630.........640.........650.........660.........670.........680
## .........690.........700.........710.........720.........730.........740........
## .750.........760.........770.........780.........790.........800.........810......
## ...820.........830.........840.........850.........860.........870.........880....
## .....890.........900.........910.........920.........930.........940.........950..
## .......960.........970.........980.........990........ 999.
## 
## Done.

plot(k_lavender_home.csr, . - r ~ r, xlab="d", ylab="K(d)-r",  main="CSR test on Full home/apt in Lavender")

As the observed k-value curve lies above the envelope, the null hypothesis is rejected and we can conclude that the distribution of Entire home/apt listings in Lavender are not randomly distributed with 99.9% confidence. Furthermore, as the observed K function curve is above the envelope, the distribution of “Entire home/apt” listings on Airbnb in Lavender subzone exhibit clustering. As the curve is not extremely steep, there are no significant cluster locations in Lavender, indicating that most of the properties are in close proximity to one another as seen in the map as well.

Distribution of Private rooms in Lavender

k_lavender_proom = Kest(lavender_proom_ppp, correction = "Ripley")
plot(k_lavender_proom, .-r ~ r, ylab= "K(d)-r", xlab = "d(m)", main="K-function plot of Private room in lavender")

CSR test on Private Rooms in Lavender

The test hypothesis is as follows:

Null hypothesis: The distribution of Private room listings in Lavender are randomly distributed.

Alternate hypothesis: The distribution of Private room listings in Lavender are not randomly distributed.

The null hypothesis will be rejected if p-value is smaller than alpha value of 0.001.

k_lavender_proom.csr <- envelope(lavender_proom_ppp, Kest, nsim = 999, nrank = 1)

## Generating 999 simulations of CSR  ...
## 1, 2, 3, ......10.........20.........30.........40.........50.........60........
## .70.........80.........90.........100.........110.........120.........130......
## ...140.........150.........160.........170.........180.........190.........200....
## .....210.........220.........230.........240.........250.........260.........270..
## .......280.........290.........300.........310.........320.........330.........340
## .........350.........360.........370.........380.........390.........400........
## .410.........420.........430.........440.........450.........460.........470......
## ...480.........490.........500.........510.........520.........530.........540....
## .....550.........560.........570.........580.........590.........600.........610..
## .......620.........630.........640.........650.........660.........670.........680
## .........690.........700.........710.........720.........730.........740........
## .750.........760.........770.........780.........790.........800.........810......
## ...820.........830.........840.........850.........860.........870.........880....
## .....890.........900.........910.........920.........930.........940.........950..
## .......960.........970.........980.........990........ 999.
## 
## Done.

plot(k_lavender_proom.csr, . - r ~ r, xlab="d", ylab="K(d)-r",  main="CSR test on Private rooms in Lavender")

As the observed k-value curve lies above the envelope, the null hypothesis is rejected and we can conclude that the distribution of Private room listings in Lavender are not randomly distributed with 99.9% confidence. Furthermore, as the observed K function curve is above the envelope, the distribution of Private rooms on Airbnb in Lavender subzone significantly exhibit clustering.

Distribution of Shared room in Aljunied

k_lavender_sroom = Kest(lavender_sroom_ppp, correction = "Ripley")
plot(k_lavender_sroom, .-r ~ r, ylab= "K(d)-r", xlab = "d(m)", main="K-function plot of Shared Room in lavender")

CSR Test

The test hypothesis is as follows:

Null hypothesis: The distribution of Shared room listings in Lavender are randomly distributed.

Alternate hypothesis: The distribution of Shared room listings in Lavender are not randomly distributed.

The null hypothesis will be rejected if p-value is smaller than alpha value of 0.001.

k_lavender_sroom.csr <- envelope(lavender_sroom_ppp, Kest, nsim = 999, nrank = 1)

## Generating 999 simulations of CSR  ...
## 1, 2, 3, ......10.........20.........30.........40.........50.........60........
## .70.........80.........90.........100.........110.........120.........130......
## ...140.........150.........160.........170.........180.........190.........200....
## .....210.........220.........230.........240.........250.........260.........270..
## .......280.........290.........300.........310.........320.........330.........340
## .........350.........360.........370.........380.........390.........400........
## .410.........420.........430.........440.........450.........460.........470......
## ...480.........490.........500.........510.........520.........530.........540....
## .....550.........560.........570.........580.........590.........600.........610..
## .......620.........630.........640.........650.........660.........670.........680
## .........690.........700.........710.........720.........730.........740........
## .750.........760.........770.........780.........790.........800.........810......
## ...820.........830.........840.........850.........860.........870.........880....
## .....890.........900.........910.........920.........930.........940.........950..
## .......960.........970.........980.........990........ 999.
## 
## Done.

plot(k_lavender_sroom.csr, . - r ~ r, xlab="d", ylab="K(d)-r",  main="CSR test on Shared rooms in Lavender")

As the observed k-value curve lies above the envelope, the null hypothesis is rejected and we can conclude that the distribution of Shared room listings in Lavender are not randomly distributed with 99.9% confidence. Furthermore, as the observed K function curve is above the envelope, the distribution of "Shared rooms on Airbnb in Lavender subzone significantly exhibit clustering.

Distribution of Hotel room in lavender

k_lavender_hotel = Kest(lavender_hotel_ppp, correction = "Ripley")
plot(k_lavender_hotel, .-r ~ r, ylab= "K(d)-r", xlab = "d(m)", main="K-function plot of Hotel Room in lavender")

CSR test

The test hypothesis is as follows:

Null hypothesis: The distribution of Hotel room listings in Lavender are randomly distributed.

Alternate hypothesis: The distribution of Hotel room listings in Lavender are not randomly distributed.

The null hypothesis will be rejected if p-value is smaller than alpha value of 0.001.

k_lavender_hotel.csr <- envelope(lavender_hotel_ppp, Kest, nsim = 999, nrank = 1, glocal=TRUE)

## Generating 999 simulations of CSR  ...
## 1, 2, 3, ......10.........20.........30.........40.........50.........60........
## .70.........80.........90.........100.........110.........120.........130......
## ...140.........150.........160.........170.........180.........190.........200....
## .....210.........220.........230.........240.........250.........260.........270..
## .......280.........290.........300.........310.........320.........330.........340
## .........350.........360.........370.........380.........390.........400........
## .410.........420.........430.........440.........450.........460.........470......
## ...480.........490.........500.........510.........520.........530.........540....
## .....550.........560.........570.........580.........590.........600.........610..
## .......620.........630.........640.........650.........660.........670.........680
## .........690.........700.........710.........720.........730.........740........
## .750.........760.........770.........780.........790.........800.........810......
## ...820.........830.........840.........850.........860.........870.........880....
## .....890.........900.........910.........920.........930.........940.........950..
## .......960.........970.........980.........990........ 999.
## 
## Done.

plot(k_lavender_hotel.csr, . - r ~ r, xlab="d", ylab="K(d)-r",  main="CSR test on Hotel rooms in Lavender")

As the observed k-value curve lies above the envelope, the null hypothesis is rejected and we can conclude that the distribution of Hotel room listings in Lavender are not randomly distributed with 99.9% confidence. Furthermore, as the observed K function curve is above the envelope, the distribution of Hotel rooms on Airbnb in Lavender subzone significantly exhibit clustering.

Insights from CSR tests in Lavender

From the CSR tests conducted above, all the room types exhibit significant clustering. However, a point to consider is that the number of shared rooms and hotel rooms are much lesser as compared to the number of full home/apt and private room listings which maybe one of the external factors contributing to the conclusions made as the K-function is very sensitive to the number of observations.

To analyse these patterns further, kernel density maps are drawn for each Room type.

Kernel density map by room type in Lavender

par(mfrow=c(2,2))
a <- as(lavender,"Spatial")
lavender_owin <- as(a,"owin")

lavender_home_ppp <- lavender_home_ppp[lavender_owin]
lavender_home_ppp.km <- rescale(lavender_home_ppp, 1000, "km")
kde_lavender_home <- density(lavender_home_ppp.km, sigma=0.25, edge=TRUE, kernel="gaussian")
kde_lavender_home_unscalled <- density(lavender_home_ppp, sigma=bw.scott, edge=TRUE, kernel="gaussian")
plot(kde_lavender_home)


lavender_proom_ppp <- lavender_proom_ppp[lavender_owin]
lavender_proom_ppp.km <- rescale(lavender_proom_ppp, 1000, "km")
kde_lavender_proom <- density(lavender_proom_ppp.km, sigma=0.25, edge=TRUE, kernel="gaussian")
plot(kde_lavender_proom)
kde_lavender_proom_unscalled <- density(lavender_proom_ppp, sigma=bw.scott, edge=TRUE, kernel="gaussian")

lavender_sroom_ppp <- lavender_sroom_ppp[lavender_owin]
lavender_sroom_ppp.km <- rescale(lavender_sroom_ppp, 500, "km")
kde_lavender_sroom <- density(lavender_sroom_ppp.km, sigma=0.25, edge=TRUE, kernel="gaussian")
plot(kde_lavender_sroom)
kde_lavender_sroom_unscalled <- density(lavender_sroom_ppp, sigma=bw.scott, edge=TRUE, kernel="gaussian")

lavender_hotel_ppp <- lavender_hotel_ppp[lavender_owin]
lavender_hotel_ppp.km <- rescale(lavender_hotel_ppp, 500, "km")
kde_lavender_hotel <- density(lavender_hotel_ppp.km, sigma=0.25, edge=TRUE, kernel="gaussian")
plot(kde_lavender_hotel)

kde_lavender_hotel_unscalled <- density(lavender_hotel_ppp, sigma=bw.scott, edge=TRUE, kernel="gaussian")

As seen in all the Kernel density maps above, a clear pattern can be observed. Full apartments/home and shared rooms are concentrated in the northern region of the subzone, wheras private and hotel rooms are concentrated in the eastern region of the subzone.

To determine the cause of this relationship, the kernel density map will be converted into a raster layer and then plotted into a base layer of OpenStreetMap to see the how various spatial elements in Lavender affect where the listings are present.

Data conversion to raster and assigning it projection

kde_lavender_home_raster <- raster(kde_lavender_home_unscalled)
projection(kde_lavender_home_raster) <- CRS("+init=EPSG:3414")
values(kde_lavender_home_raster) <- values(kde_lavender_home_raster)*100000

kde_lavender_proom_raster <- raster(kde_lavender_proom_unscalled)
projection(kde_lavender_proom_raster) <- CRS("+init=EPSG:3414")
values(kde_lavender_proom_raster) <- values(kde_lavender_proom_raster)*100000

kde_lavender_sroom_raster <- raster(kde_lavender_sroom_unscalled)
projection(kde_lavender_sroom_raster) <- CRS("+init=EPSG:3414")
values(kde_lavender_sroom_raster) <- values(kde_lavender_sroom_raster)*100000

kde_lavender_hotel_raster <- raster(kde_lavender_hotel_unscalled)
projection(kde_lavender_hotel_raster) <- CRS("+init=EPSG:3414")
values(kde_lavender_hotel_raster) <- values(kde_lavender_hotel_raster)*100000

Visualising the raster data on openstreemap

tmap_mode("view")
tm_shape(kde_lavender_home_raster)+
  tm_raster(colorNA = "lightgrey", showNA = FALSE,alpha=0.5,title = "Number of listings")+
  tm_shape(lavender)+
  tm_borders(lwd = 2)+
  tm_basemap(group = "OpenStreetMap")+
  tm_layout("Distribution of Full home/apt in lavender")

tm_shape(kde_lavender_proom_raster)+
  tm_raster(colorNA = "lightgrey", showNA = FALSE,alpha=0.5,title = "Number of listings")+
  tm_shape(lavender)+
  tm_borders(lwd = 2)+
  tm_basemap(group = "OpenStreetMap")+
  tm_layout("Distribution of Private room in lavender")

tm_shape(kde_lavender_sroom_raster)+
  tm_raster(colorNA = "lightgrey", showNA = FALSE,alpha=0.5,title = "Number of listings")+
  tm_shape(lavender)+
  tm_borders(lwd = 2)+
  tm_basemap(group = "OpenStreetMap")+
  tm_layout("Distribution of Shared room in lavender")

tm_shape(kde_lavender_hotel_raster)+
  tm_raster(colorNA = "lightgrey", showNA = FALSE,alpha=0.5,title = "Number of listings")+
  tm_shape(lavender)+
  tm_borders(lwd = 2)+
  tm_basemap(group = "OpenStreetMap")+
  tm_layout("Distribution of Hotel room in lavender")
tmap_mode("plot")

Screenshots of maps

Interpretation

Most of the Full apt/home listings are located in the northern region of Lavender in close proximity of Farrer Park MRT. This localilty has more Full apt/home listings as compared to the other parts of Lavender as most of the private condominiums are located here, with the added proximity of shopping centres which makes this a popular location for an Airbnb listing. This location is also where most of the hotels in the region are located due to its convinience. Furthermore, this location is also where most of the private rooms are located. Private rooms are also located in the centre of Lavender, by Jalan Besar Road near the Jalan Besar MRT station. It is very evident that most of the listings are located in close proximity of public transportation facilities as compared to locations which are distant from public transport facilities. Most of the airbnb listings are located in the northern part of Lavender, due to its convinience and also located near the infamous Little India.

Visualising Airbnb listing by room type in Tanjong Pagar

data <- st_intersection(airbnb3414,tanjongpagar)
data_tp <- st_jitter(data)
tmap_mode("plot")

## tmap mode set to plotting

tm_shape(tanjongpagar)+
  tm_borders()+
  tm_shape(data_tp)+
  tm_facets(by="room_type",free.coords = FALSE)+
  tm_dots("room_type",size=0.5,alpha=0.7,title = "Room type")+
  tm_layout(main.title="Distribution of Airbnb listings by room type in Tanjong Pagar")

We will now examine the spatial patterns observed in each of the distributions. Firstly, there are no Shared rooms in Tanjong Pagar, hence there is no plot drawn for that. Entire home/apartment listings are present in all regions but concentrated towards the south-western region of Tanjong Pagar.Private r are located in nearly all the parts of Tanjong Pagar. There are only two hotel room listings in Tanjong Pagar, hence we won’t be performing any spatial analysis on it.

To gain statistical understanding of the distribution, K function is performed on every room type except shared room and hotel room to determine if there is complete spatial randomness. ### Data preperation for every room type

tp_home <- data_tp%>%filter(room_type=="Entire home/apt")
a<- as(tp_home,"Spatial")
b <- as(a,"SpatialPoints")
tp_home_ppp <- as(b,"ppp")

tp_proom <- data_tp%>%filter(room_type=="Private room")
a<- as(tp_proom,"Spatial")
b <- as(a,"SpatialPoints")
tp_proom_ppp <- as(b,"ppp")

Second order spatial analysis in Tanjong Pagar through K function

Distribution of Entire home/apt in Tanjong Pagar

k_tp_home = Kest(tp_home_ppp, correction = "Ripley")
plot(k_tp_home, .-r ~ r, ylab= "K(d)-r", xlab = "d(m)", main="K-function plot of Entire home/apt in lavender")

CSR test

The test hypothesis is as follows:

Null hypothesis: The distribution of Entire home/apt listings in Tanjong Pagar are randomly distributed.

Alternate hypothesis: The distribution of Entire home/apt listings in Tanjong Pagar are not randomly distributed.

The null hypothesis will be rejected if p-value is smaller than alpha value of 0.001.

k_tp_home.csr <- envelope(tp_home_ppp, Kest, nsim = 999, nrank = 1)

## Generating 999 simulations of CSR  ...
## 1, 2, 3, ......10.........20.........30.........40.........50.........60........
## .70.........80.........90.........100.........110.........120.........130......
## ...140.........150.........160.........170.........180.........190.........200....
## .....210.........220.........230.........240.........250.........260.........270..
## .......280.........290.........300.........310.........320.........330.........340
## .........350.........360.........370.........380.........390.........400........
## .410.........420.........430.........440.........450.........460.........470......
## ...480.........490.........500.........510.........520.........530.........540....
## .....550.........560.........570.........580.........590.........600.........610..
## .......620.........630.........640.........650.........660.........670.........680
## .........690.........700.........710.........720.........730.........740........
## .750.........760.........770.........780.........790.........800.........810......
## ...820.........830.........840.........850.........860.........870.........880....
## .....890.........900.........910.........920.........930.........940.........950..
## .......960.........970.........980.........990........ 999.
## 
## Done.

plot(k_tp_home.csr, .-r ~ r, ylab= "K(d)-r", xlab = "d(m)", main="K-function plot of Entire home/apt in Tanjong Pagar")

As the observed k-value curve lies above the envelope, the null hypothesis is rejected and we can conclude that the distribution of Entire home/apt listings in Tanjong Pagar are not randomly distributed with 99.9% confidence. Furthermore, as the observed K function curve is above the envelope, the distribution of “Entire home/apt” listings on Airbnb in Tanjong Pagar subzone exhibit clustering.

Distribution of Private room in lavender

k_tp_proom = Kest(tp_proom_ppp, correction = "Ripley")
plot(k_tp_proom, .-r ~ r, ylab= "K(d)-r", xlab = "d(m)", main="K-function plot of Private rooms in Tanjong Pagar")

As seen in the above graph, as the K-function value is over its theoretical value for some distances and below its theoritical distance, we will perform a complete spatial randomness test to examine if there are signs of uniformity or clustered data.

CSR test

The test hypothesis is as follows:

Null hypothesis: The distribution of Private room listings in Tanjong Pagar are randomly distributed.

Alternate hypothesis: The distribution of Private room listings in Tanjong Pagar are not randomly distributed.

The null hypothesis will be rejected if p-value is smaller than alpha value of 0.001.

k_tp_proom.csr <- envelope(tp_proom_ppp, Kest, nsim = 999, nrank = 1)

## Generating 999 simulations of CSR  ...
## 1, 2, 3, ......10.........20.........30.........40.........50.........60........
## .70.........80.........90.........100.........110.........120.........130......
## ...140.........150.........160.........170.........180.........190.........200....
## .....210.........220.........230.........240.........250.........260.........270..
## .......280.........290.........300.........310.........320.........330.........340
## .........350.........360.........370.........380.........390.........400........
## .410.........420.........430.........440.........450.........460.........470......
## ...480.........490.........500.........510.........520.........530.........540....
## .....550.........560.........570.........580.........590.........600.........610..
## .......620.........630.........640.........650.........660.........670.........680
## .........690.........700.........710.........720.........730.........740........
## .750.........760.........770.........780.........790.........800.........810......
## ...820.........830.........840.........850.........860.........870.........880....
## .....890.........900.........910.........920.........930.........940.........950..
## .......960.........970.........980.........990........ 999.
## 
## Done.

plot(k_tp_proom.csr,main="K-function plot of Private room in Tanjong Pagar")

As the observed k-value curve lies above within the envelope, the null hypothesis is not rejected and we can conclude that the distribution of Private room listings in Tanjong Pagar are randomly distributed with 99.9% confidence. Therefore, the distribution of private rooms in tanjong pagar conform to spatial randomness.

Insights from CSR tests in Tanjong Pagar

Entire home/apartment are clustered in Tanjong Pagar whereas Private rooms conform to spatial randomness. However, a point to consider is that there are 149 listings for Full apartment/home whereas there are only 21 listings for Private rooms. As Tanjong Pagar is located in the heart of the CBD (Central Business District), the amount of residential area is much less as compared to he amount of area for business purposes. Therefore, given the limited area, clusters are bound to form.

To analyse these patterns further, kernel density maps are drawn for each Room type.

Kernel density maps by room type in Tanjong Pagar

a <- as(tanjongpagar,"Spatial")
tp_owin <- as(a,"owin")
par(mfrow=c(2,2))

tp_home_ppp <- tp_home_ppp[tp_owin]
tp_home_ppp.km <- rescale(tp_home_ppp, 125, "km")
kde_tp_home <- density(tp_home_ppp.km, sigma=bw.scott, edge=TRUE, kernel="gaussian")
plot(kde_tp_home, main="Kernel density map of Full apartment/home in Tanjong Pagar")
kde_tp_home_unscalled <- density(tp_home_ppp, sigma=bw.scott, edge=TRUE, kernel="gaussian")


tp_proom_ppp <- tp_proom_ppp[tp_owin]
tp_proom_ppp.km <- rescale(tp_proom_ppp, 250, "km")
kde_tp_proom <- density(tp_proom_ppp.km, sigma=bw.scott, edge=TRUE, kernel="gaussian")
plot(kde_tp_proom, main="Kernel density map of Private Room in Tanjong Pagar")
kde_tp_proom_unscalled <- density(tp_proom_ppp, sigma=bw.scott, edge=TRUE, kernel="gaussian")

Data conversion to raster and assigning it projection

par(mfrow = c(2, 1))
kde_tp_home_raster <- raster(kde_tp_home_unscalled)
projection(kde_tp_home_raster) <- CRS("+init=EPSG:3414")
values(kde_tp_home_raster) <- values(kde_tp_home_raster)*10000

kde_tp_proom_raster <- raster(kde_tp_proom_unscalled)
projection(kde_tp_proom_raster) <- CRS("+init=EPSG:3414")
values(kde_tp_proom_raster) <- values(kde_tp_proom_raster)*100000

Visualising the raster data on openstreemap

tmap_mode("view")
tm_shape(kde_tp_home_raster)+
  tm_raster(colorNA = "lightgrey", showNA = FALSE,alpha=0.5,title = "Number of listings")+
  tm_shape(tanjongpagar)+
  tm_borders(lwd = 2)+
  tm_basemap(group = "OpenStreetMap")+
  tm_layout("Distribution of Full home/apt in tp")

tm_shape(kde_tp_proom_raster)+
  tm_raster(colorNA = "lightgrey", showNA = FALSE,alpha=0.5,title = "Number of listings")+
  tm_shape(tanjongpagar)+
  tm_borders(lwd = 2)+
  tm_basemap(group = "OpenStreetMap")+
  tm_layout("Distribution of Private room in Tanjong Pagar")

Screenshots of map

Interpretation

Tanjong Pagar is a very small subzone with an area of only 0.15 km^2. It lies in the heart of the central business district and hence most buildings are for commercial or entertainment purposes. Therefore without much option of choice available, both entire home/apt and private rooms are located in the soutern part of Tanjong Pagar. Close proximity to the CBD makes it an ideal location for people who want to live close to where they work, in the heart of the city. From the map above we can see that the western part of this region contains only a handful of buildings, most of which are skyscarappers used for commercial purposes such as the AXA towers, MAS Building, etc. The northern part of Tanjong Pagar does contain residential towers, however, does not contain many Airbnb listings, possibly because of regulations.

assignment2

Amey Rathi

5/13/2020

Importing Required Packages

Importing Data

Converting airbnb to 3414 crs

Section A: Nation Wide Analysis of Airbnb Listings

Viewing all the locations of airbnb listings by house type

Performing First Order Spatial Analysis

Preparing data for each room type

Checking for duplicated data

Plotting the data

Clark and evans test for Entire home/Apartment

Clark and evans test for Hotel rooms

Clark and evans test for Private rooms

Clark and evans test for Shared rooms

Conclusion

Kernel Density Estimation

Plotting kernal density maps

Conversion of gridded objectss into raster data

Viewing airbnb listing by room type in Singapore

Interpretation and Significance of Kernel Density Maps

Section B: Analysis by planning subzones

Exploratory Data Analysis

Data preperation by extracting subzones

Visualising Airbnb listing by room type within Aljunied

Data preperation for each room type

Second order spatial analysis in Aljunied through K function

CSR test for Full home/apt in Aljunied

Distribution of Private Room in Aljunied

CSR test on Private Rooms in Aljunied

Distribution of Shared Room in Aljunied

Distribution of Hotel Room in Aljunied

Insights from CSR tests in Aljunied

Kernel density maps by room type in Aljunied

Data conversion to raster and assigning it projection

Visualising the raster data on openstreemap

Screenshots of maps

Interpretation

Visualising Airbnb listing by room type within Balestier

Data preperation for every room type

Second order spatial analysis in Balestier through K function

Distribution of Full home/apt in Balestier through K function

Performing CSR test

Distribution of Hotel room in Balestier

CSR test

Distribution of Private room in Balestier

CSR test

Data preperation

Plotting Kernel density

Data conversion to raster and assigning it projection

Visualising the raster data on openstreemap

Screenshots of maps

Interpretation

Visualising Airbnb listing by room type within Lavender

Data preperation for each room type

Second order spatial analysis in Lavender through K function

Distribution of Entire home/apt in Lavender

CSR test

Distribution of Private rooms in Lavender

CSR test on Private Rooms in Lavender

Distribution of Shared room in Aljunied

CSR Test

Distribution of Hotel room in lavender

CSR test

Insights from CSR tests in Lavender

Kernel density map by room type in Lavender

Data conversion to raster and assigning it projection

Visualising the raster data on openstreemap

Screenshots of maps

Interpretation

Visualising Airbnb listing by room type in Tanjong Pagar

Second order spatial analysis in Tanjong Pagar through K function

Distribution of Entire home/apt in Tanjong Pagar

CSR test

Distribution of Private room in lavender

CSR test

Insights from CSR tests in Tanjong Pagar

Kernel density maps by room type in Tanjong Pagar

Data conversion to raster and assigning it projection