Locate your Data and Boost it with Geo-Processing Post

geoprocess_dd_post.py

# Getting Latitude and Longitude from Nominatim

from geopy.geocoders import Nominatim
from geopy.extra.rate_limiter import RateLimiter

geocoder = Nominatim(user_agent="FindAddress")
geocode = RateLimiter(
    geocoder.geocode,
    min_delay_seconds = 1,
    return_value_on_exception = None
)


# Returns the location from the Nominatim server
geocode("Areeiro, Lisboa, Portugal")

# Using GeoPandas

import geopandas as gpd

house_data_gdf = gpd.GeoDataFrame(
    house_data,
    geometry=gpd.points_from_xy(
       house_data.longitude,
       house_data.latitude
    ),
    crs="epsg:4326",
)

# Changing the CRS to a Projected Coordinate System ideal for Portugal.
house_data_gdf.to_crs(epsg=3857, inplace=True)

# Obtaining Shape File for Portugal
parishes_url = "zip+https://dados.gov.pt/s/resources/freguesias-de-portugal/20181112-195834/cont-aad-caop2017.zip"
parishes = gpd.read_file(parishes_url)
parishes.head()

# Joining House Data with Parish
gpd.sjoin(house_data_gdf, parishes, how="left", op="within")

# Counting the Hospitals near Houses
house_data_gdf_buffer = (
   house_data_gdf
   .copy()
   .assign(geometry_buffer = lambda d: d.buffer(1000))
   .set_geometry("geometry_buffer")
)
 
# To get the a statistic of a geometry within another geometry, 
# we can use a spatial join and then aggregate the values
house_data_gdf["hospitals_in_1km"] = (
    gpd.sjoin(
        house_data_gdf_buffer,
        hospitals_gdf,
        how="left",
        op="contains"
   )
   .groupby("house_id", as_index=False)
   .agg({"OBJECTID": "count"})
   .rename(columns={"OBJECTID":"Hospitals"})
   .loc[:, "Hospitals"]
)

# Calculating Distance from Metro to each House
# Read the data
metro = pd.read_csv(Path("data", "metro_stations.csv"))

# Convert to a GeoDataFrame
metro = gpd.GeoDataFrame(
    metro,
    geometry=gpd.points_from_xy(metro.longitude, metro.latitude),
    crs="epsg:4326"
).to_crs(epsg=3857)

# Get the closest metro station to each house
closest_metros = gpd.sjoin_nearest(
    house_data_gdf,
    metro,
    how="left"
).loc[:, ["latitude_right", "longitude_right"]]

# Convert the closest metro location to a geometry
house_data_gdf["closest_metro"] = gpd.points_from_xy(
    closest_metros.longitude_right,
    closest_metros.latitude_right,
    crs="epsg:4326"
).to_crs(epsg=3857)

# Calculate the distance to the closest metro station
house_data_gdf.assign(
    metro_distance=house_data_gdf.distance(
        house_data_gdf.closest_metro, align=True))

# Obtaining Elevation
import rasterio
from rasterio.plot import show
 
url = "zip+file:data/mdt.zip!mdt.tif"
lisbon_elevation = rasterio.open(url)

# Plot the raster data to get a sense of it
show(lisbon_elevation, cmap="terrain")

# Get the elevation from the raster data
house_data_gdf["elevation"] = (
    house_data_gdf
    .to_crs(lisbon_elevation.crs)
    .geometry
    .apply(
        lambda x: next(lisbon_elevation.sample([(x.x, x.y)]))[0]
    )
)

# Performing statistics
y = house_data_gdf['price_per_m2'].values
w = KNN.from_dataframe(house_data_gdf, k=1000)
w.transform = "r"

from splot.esda import lisa_cluster
from esda.moran import Moran_Local
import contextily as ctx
 
moran_loc = Moran_Local(y, w)
fig, ax = lisa_cluster(moran_loc, gdf, p=0.05, figsize = (10,10), markersize = 500)
ctx.add_basemap(ax, source=ctx.providers.Stamen.TonerLite)
plt.show()