Reuters

#30DayChartChallenge Day 23

In today’s blog post, we’ll explore the concept of “Tiles” under the broader category of “Timeseries.” I wanted to revisit the usage of geom_tiles, which often creates heatmap-like charts. One interesting aspect of heatmaps is that they provide valuable insights, especially when one of the axes is ordinal, and even more so if it’s a time axis.

I was very inspired by this guide that depicted the declining incidence of measles in the US over the years.

After a bit of a search, I found a dataset from data.gov.sg that covered a wide time period tracking Singapore’s average resale housing prices. Unsurprisingly, the data revealed an increasing trend in housing prices over time. However, there are still a few intriguing insights to be gleaned regarding which neighbourhoods have always had higher resale prices, and which neighbourhoods only saw a more recent rise.

import polars as pl
import polars.selectors as cs
import datetime
from lets_plot import *

LetsPlot.setup_html()

# Data Cleaning
raw = (pl.read_csv('Resale_Flat_prices.csv',ignore_errors=True)
    .with_columns(
        pl.col('month').str.to_date('%Y-%m').alias('date'),
        pl.col('month').str.to_date('%Y-%m').dt.year().alias('year'),
    )
)

df = (raw
    .group_by(['month','date','town'])
    .agg(pl.col('resale_price').mean())
    .sort(by=['date','town'], descending=[False, True])
)


(ggplot(df, aes(x='date',y='town',fill='resale_price'))
    +geom_tile(size=0.8)
    +scale_x_discrete(
        labels=[str(date.year) if date.month == 1 else " " for date in df["date"].unique().to_list()]
    )
    +labs(
        x="",
        y="",
        title="The rise of Singapore Housing's Average Resale Prices",
        subtitle="From 2017 to 2024, across major neighbourhoods",
        caption = '#30DayChartChallenge #Day23 Tiles\nData: data.gov.sg\nMade by: www.ddanieltan.com',
        fill='Resale Price (SGD)'
    )
    + theme(
        axis_text_x = element_text(angle=1, size=10),
        axis_text_y = element_text(size=5),
        legend_position = 'right',
        legend_text= element_text(size=8),
        legend_title= element_text(size=10),
        plot_subtitle = element_text(size=10),
        plot_caption=element_text(size=12, color='grey'),
    )
    + scale_fill_brewer(palette='YlGnBu')
    + ggsize(1000,800)
)

TIL

1. In polars, when selecting a column using df.select(<col>) a Polars DataFrame is returned, and the DataFrame does not have a to_list() method. However, when selecting a column using df["<col>"] a Polars Series is returned that does have a to_list() method.

2. I faced a lot of difficulty in specifying which ticks I want to appear on the chart for my x axis of type datetime.date. My eventual solution seemed a little hacky but I’m satisfied I got close to what I wanted to visualise.