Singapore HDB’s Mean Resale Price in 2019

Key Challenges

1. Data Challenge - Missing Year Data

The source dataset is missing the year data. The date field has concatenate both the year and month information together, in the form 2017-01, 2017-02, etc. The challenge would be how do we extract the year data from this format.

2. Design Challenge - No obvious cluster for visualization

There are a total of 26 towns. Visualizing this dataset would be difficult. It is not feasible to use the 26 towns as a cluster, because it would be hard for users to understand what is being visualized. It would also be hard for users to draw valuable insights if we used a huge cluster size.

3. Design Challenge - No obvious common field between aspatial and geospatial dataset

Our aspatial resale data location information is ‘town’, there isn’t a field ‘town’ in our geospatial dataset. Given the geospatial dataset has many locational field, the challenge would be mapping common field correctly.

Suggestions to Overcome Challenges

Proposed Design

Step 1: Load Library

packages = c('sf', 'tmap', 'tidyverse', 'highcharter', 'plotly')
for (p in packages){
  if(!require(p, character.only = T)){
    install.packages(p)
  }
  library(p,character.only = T)
}

Step 2: Load Data

resale <- read.csv("C:/Users/think/Desktop/1-SMU Term 3/5-Visual Analytics and Applications/Assignment 5/data/aspatial/resale.csv", header = T)

mpsz <- st_read(dsn = "C:/Users/think/Desktop/1-SMU Term 3/5-Visual Analytics and Applications/Assignment 5/data/geospatial", layer = "MP14_SUBZONE_WEB_PL")

## Reading layer `MP14_SUBZONE_WEB_PL' from data source `C:\Users\think\Desktop\1-SMU Term 3\5-Visual Analytics and Applications\Assignment 5\data\geospatial' 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
## projected CRS:  SVY21

Step 3: Extract 2019 dataset and important fields

resale2019 <- resale %>%
  filter(year == 2019) %>%
  select('town','flat_type','storey_range', 'remaining_lease', 'resale_price')

Step 4: Visualize the mean resale price broken down by town, in an interactive bar chart

First, prepare the data using this code:

options(highcharter.theme = hc_theme_smpl(tooltip = list(valueDecimals = 2)))

plotmean <- resale2019 %>%
  group_by(town) %>% 
  summarise(Mean_Resale_Price = mean(resale_price))

plotmean_sorted = plotmean[order(plotmean$Mean_Resale_Price),]

Next, plot the interactive barchart using this code:

hc <- plotmean_sorted %>%
  hchart('column', hcaes(x = town, y = Mean_Resale_Price)) %>%
  hc_title(text="HDB Mean Selling Price in 2019", align = "center")

Step 5: Visualize whether higher storey HDB commands a higher price, in an interactive heatmap

First, prepare the data using this code:

plotfloor<- resale2019 %>%
  group_by(town, storey_range) %>% 
  summarise(Mean_Resale_Price = mean(resale_price)) 

Next, plot the interactive heatmap using this code:

heatmap <- ggplot(data = plotfloor, 
                  mapping = aes(x = town, y = storey_range,fill = Mean_Resale_Price)) +
            geom_tile() +
  labs(title = "Heatmap of HDB breakdown by area and storey", x = "Town", y = "Storey") +
  scale_fill_gradient(name = "Mean Resale Price",
                      low = "#ffedde",
                      high = "#c24b0c")+
  theme(axis.text.x = element_text(angle = 45))

Step 6: Visualize the mean resale price using a Choropleth Map

First, prepare the data using this code:

mpsz_resale2019 <- left_join(mpsz, plotmean, by = c("PLN_AREA_N" = "town"))
tmap_mode("view")

Next, plot an interactive choropleth map using this code:

tm_shape(mpsz_resale2019)+
  tm_fill("Mean_Resale_Price",
          n = 6,
          style = "quantile", 
          palette = "Oranges") +
  tm_borders(alpha = 0.5)

For deeper analysis, we further broken down the choropleth map into different regions using this code:

tmap_options(limits=c(facets.plot=32, facets.view=5))

tm_shape(mpsz_resale2019)+
  tm_fill("Mean_Resale_Price",
          n = 6,
          style = "quantile", 
          palette = "Oranges") +
  tm_borders(alpha = 0.5)+
  tm_facets(by="REGION_N", free.coords=TRUE)+
  tm_layout(legend.show = FALSE) +
  tm_borders(alpha = 0.5)

Insight 1: HDB in Bukit Timah, Bishan and Central Area are the most expensive

Insight 2.1: Higher storey command better prices

Insight 2.2: There is no price differentiation if you stay at unpopular town

Insight 2.3: Bukit Timah does not have high storey building, while Central has the most high storey building

First, based on the heatmap, it is generally true that higher storey units commands a better price, as illustrated from the darker colours on higher storey.

However, if you stay in unpopular towns such as Woodlands, Sembawang, Choa Chu Kang, Jurong West, Pasir Ris, Punggol and Hougang, there is not much differentiation in price even for higher storey units. This is probably because these town are at the edges of Singapore. It is extremely inconvenient to commute from those areas to town. Therefore, there is less demand for those flats resulting in a lower price.

Finally, from the heatmap, it is observed that flats in Bukit Timah is relatively low rised compared to other towns. The highest storey in Bukit Timah is only 15. This is probably because Bukit Timah is known as an exclusive area with many Bungalows and Terraces houses. Therefore, to maintain its “prestige”, the urban planning ministry might have intentionally kept HDB storey at Bukit Timah low too.

## `summarise()` regrouping output by 'town' (override with `.groups` argument)

Insight 3: HDB in central region are more expensive, while North region is the cheapest

Based on the choropleth map below, it is observed that certain areas are more expensive than others. Prices are usually similar for the surrounding cluster.

When we dive deeper into the map and classify them into their respective region, it is observed that HDB in central area are the most expensive.

HDB in central region such as Toa-Payoh, Kallang and Tanjong Pagar, are usually mature towns with many amenities, shopping centers, offices and gym facilities. Therefore it is reasonable that they command a higher price, for the convenience the location provides.

It is also observed that HDB in the North region are the cheapest, as these are usually new towns like Seng Kang with fewer amenities.

## tmap mode set to plotting