Skip to content

Latest commit

 

History

History
68 lines (67 loc) · 5.24 KB

README.md

File metadata and controls

68 lines (67 loc) · 5.24 KB
Raw

Data science for real estate

Built Heritage vs Real Estate Market. The Barcelona case

BLog link
Data science project in Python
tools: Jupyter Notebook, Python, Idealista API + Pandas + Geopandas + Numpy + Folium + BeautifulSoup + OSMNX + Matplotlib + Plotly + Networkx + Scikit-learn.

Questions

  • How official status of built heritage affects real estate prices?
  • How correlation could be identified geographically?
  • What is the difference between correlations of oficially listed built heritage or old buildings over 50 years with real estate prices?

What I learned:

  • Idealista API for query of real estate lots
  • Geographical distribution of data by natural jenks
  • Geographical distribution of data by hexagon histogram
  • Make comprehensive maps with Folium
  • Measuring regions the Spatial lag
  • Clustering maps by K-Means
  • Merging and overlapping maps built from different datasets for correlation insights

Key takeaways

  • spatial affect of built heritage spreads within isochrones of 10-munites walk from listed heritage buildings
  • strong correlation observed not only between the real estate cost and alone standing heritage buildings but also between the real estate cost and historical neighborhoods that include not-listed old buildings.
  • the spatial lag reveals the highest demand for real estate in neighborhoods La Dreta and Les Cortes.
  • 50% of all residential buildings in Barcelona were built between 1920-1970s, and together with more oldest 22% they combine the 72% cogort of buildings older 50 years.

video presentation

ALt text

methodology

  1. Evaluate the best way of data mining in order to get sufficent data for analysis. source>>
  2. Calculate the price per sqm for each data entry in the real estate prices dataset.source>>
  3. Match the real estate prices dataset with the Barcelona neighborhood shapefile, calculate mean, max, min, and average price per sqm. source>>
  4. Fetch heritage points and polygons from OpenStreetMap, create isochrones of them at 100, 400, and 800 meters. source>>
  5. Fetch buildings data from the public cadastre, categorize them by heritage-like and not (already 50 years old buildings could fit the built heritage policy requirements), cluster buildings, make buffers of 400m, match the amount of heritage-like buildings with the Barcelona neighborhood shapefile. source>>
  6. Make a correlation between mean, max, min, and average price per sqm and heritage buffer zones, heritage-like buffer zones. source>>
  7. Evaluate the hypothesis, check the data on conditions of heritage buildings in non-correlated zones. source>>

data visualization

Map representing geographical distribution of listings retrieved by Idealista API and aggregated by natural jenks

Alt text

Map representing geographical distribution of listings retrieved by ready dataset sourced from Fotocasa and aggregated by natural jenks

Hexagon histogram from the previous geodataset

GIF Map visualization of the merged dataset of listings and Barcelona neighborhoods showing amount of listings, minimum, maximum, average, mean, medium prices.

Map data visualization of the spatial lag of mean prices by neighborhoods of Barcelona.

Map of data visualization of listings clustered by price. K-means by Scikit-learn

Lollipop graph of clusters.

Map data visualization of overlapping ORANGE 100-m buffers around osm heritage amentites and BLUE 2 heighest clusters of listings.

Map data visualization of 5 - 10 - 20 min isochrones around osm heritage amentites.

Map data visualization the histogram range by the mean apartment price within heritage 10-min isochrone areas.

Map data visualization the buildings footprints clustered by 5 ranges of construction years.

Donut chart of 5 clusters of buildings in Barcelona by age.

Map data visualization representing 5-age clustering the all residential and heritage-listed buildings.

Map data visualization of overlapping datasets of over-100-years-old buildings and 10-min walkability isochrones of 2 heighest price range clusters of listings.

References

  • OpenStreetMap: OpenStreetMap
  • Institut Cartogràfic i Geològic de Catalunya: ICGC
  • Public Cadastre of Spain: Cadastre
  • Idealista.es – Spain's #1 Real Estate Web Portal: Idealista
  • Fotocasa.es – Spain's #2 Real Estate Web Portal: Fotocasa