Reclassifying data#

Reclassifying data based on specific criteria is a common task when doing GIS analysis. The purpose of this lesson is to see how we can reclassify values based on some criteria. We could, for example, classify information based on travel times and housing prices using these criteria:

  1. if travel time to my work is less than 30 minutes, AND

  2. the rent of the apartment is less than 1000 € per month

If both criteria are met: I go to see the appartment and try to rent it If not: I continue to look for something else

In this tutorial, we will:

  1. Use classification schemes from the PySAL mapclassify library to classify travel times into multiple classes.

  2. Create a custom classifier to classify travel times and distances in order to find out good locations to buy an apartment with these conditions:

    • good public transport accessibility to city center

    • bit further away from city center where the prices are presumably lower

Input data#

We will use Travel Time Matrix data from Helsinki that contains travel time and distance information for routes between all 250 m x 250 m grid cell centroids (n = 13231) in the Capital Region of Helsinki by walking, cycling, public transportation and car.

import pathlib 
NOTEBOOK_PATH = pathlib.Path().resolve()
DATA_DIRECTORY = NOTEBOOK_PATH / "data"
import geopandas
accessibility_grid = geopandas.read_file(
    DATA_DIRECTORY
    / "helsinki_region_travel_times_to_railway_station"
    / "helsinki_region_travel_times_to_railway_station.gpkg"
)
accessibility_grid.head()
car_m_d car_m_t car_r_d car_r_t from_id pt_m_d pt_m_t pt_m_tt pt_r_d pt_r_t pt_r_tt to_id walk_d walk_t geometry
0 32297 43 32260 48 5785640 32616 116 147 32616 108 139 5975375 32164 459 POLYGON ((382000.000 6697750.000, 381750.000 6...
1 32508 43 32471 49 5785641 32822 119 145 32822 111 133 5975375 29547 422 POLYGON ((382250.000 6697750.000, 382000.000 6...
2 30133 50 31872 56 5785642 32940 121 146 32940 113 133 5975375 29626 423 POLYGON ((382500.000 6697750.000, 382250.000 6...
3 32690 54 34429 60 5785643 33233 125 150 33233 117 144 5975375 29919 427 POLYGON ((382750.000 6697750.000, 382500.000 6...
4 31872 42 31834 48 5787544 32127 109 126 32127 101 121 5975375 31674 452 POLYGON ((381250.000 6697500.000, 381000.000 6...

Common classifiers#

Classification schemes for thematic maps#

PySAL -module is an extensive Python library for spatial analysis. It also includes all of the most common data classifiers that are used commonly e.g. when visualizing data. Available map classifiers in pysal’s mapclassify -module:

  • Box Plot

  • Equal Interval

  • Fisher Jenks

  • Fisher Jenks Sampled

  • HeadTail Breaks

  • Jenks Caspall

  • Jenks Caspall Forced

  • Jenks Caspall Sampled

  • Max P Classifier

  • Maximum Breaks

  • Natural Breaks

  • Quantiles

  • Percentiles

  • Std Mean

  • User Defined

There are plenty of different variables in the accessibility data set (see from here the description for all attributes) but what we are interested in are columns called pt_r_tt which is telling the time in minutes that it takes to reach city center from different parts of the city, and walk_d that tells the network distance by roads to reach city center from different parts of the city (almost equal to Euclidian distance).

The NoData values are presented with value -1. Thus we need to remove the No Data values first.

# Include only data that is above or equal to 0
accessibility_grid = accessibility_grid.loc[accessibility_grid["pt_r_tt"] >=0]

Let’s plot the data and see how it looks like

  • cmap parameter defines the color map. Read more about choosing colormaps in matplotlib

  • scheme option scales the colors according to a classification scheme (requires mapclassify module to be installed):

# Plot using 9 classes and classify the values using "Natural Breaks" classification
accessibility_grid.plot(column="pt_r_tt", scheme="Natural_Breaks", k=9, cmap="RdYlBu", linewidth=0, legend=True)
<AxesSubplot: >
../../_images/974fc2d25ce8453d01385c57e948f0cbb19a5b90cbd89ff0c63b9e6544fb6031.png

As we can see from this map, the travel times are lower in the south where the city center is located but there are some areas of “good” accessibility also in some other areas (where the color is red).

Let’s also make a plot about walking distances:

# Plot walking distance
accessibility_grid.plot(column="walk_d", scheme="Natural_Breaks", k=9, cmap="RdYlBu", linewidth=0, legend=True)
<AxesSubplot: >
../../_images/ba0893cfe8228ac58e1a4b9fa1d0afad05849105e50ae3f120906be227c5f9cf.png

Okay, from here we can see that the walking distances (along road network) reminds more or less Euclidian distances.

Applying classifiers to data#

As mentioned, the scheme option defines the classification scheme using pysal/mapclassify. Let’s have a closer look at how these classifiers work.

import mapclassify

Natural Breaks#

mapclassify.NaturalBreaks(y=accessibility_grid["pt_r_tt"], k=9)
NaturalBreaks           

    Interval       Count
------------------------
[  0.00,  31.00] |   926
( 31.00,  43.00] |  2272
( 43.00,  53.00] |  2272
( 53.00,  64.00] |  2125
( 64.00,  76.00] |  1707
( 76.00,  88.00] |  1521
( 88.00, 103.00] |  1160
(103.00, 127.00] |   838
(127.00, 199.00] |   199

Quantiles (default is 5 classes):#

mapclassify.Quantiles(y=accessibility_grid["pt_r_tt"])
Quantiles               

    Interval       Count
------------------------
[  0.00,  41.00] |  2674
( 41.00,  52.00] |  2576
( 52.00,  66.00] |  2623
( 66.00,  85.00] |  2616
( 85.00, 199.00] |  2531

Extract threshold values#

It’s possible to extract the threshold values into an array:

classifier = mapclassify.NaturalBreaks(y=accessibility_grid["pt_r_tt"], k=9)
classifier.bins
array([ 32.,  44.,  55.,  66.,  77.,  89., 103., 125., 199.])

Let’s apply one of the Pysal classifiers into our data and classify the travel times by public transport into 9 classes The classifier needs to be initialized first with make() function that takes the number of desired classes as input parameter

# Create a Natural Breaks classifier
classifier = mapclassify.NaturalBreaks.make(k=9)
  • Now we can apply that classifier into our data by using apply -function

# Classify the data
classifications = accessibility_grid[["pt_r_tt"]].apply(classifier)

# Let's see what we have
classifications.head()
pt_r_tt
0 8
1 8
2 8
3 8
4 7
type(classifications)
pandas.core.frame.DataFrame

Okay, so now we have a DataFrame where our input column was classified into 9 different classes (numbers 1-9) based on Natural Breaks classification.

We can also add the classification values directly into a new column in our dataframe:

# Rename the column so that we know that it was classified with natural breaks
accessibility_grid["nb_pt_r_tt"] = accessibility_grid[["pt_r_tt"]].apply(classifier)

# Check the original values and classification
accessibility_grid[["pt_r_tt", "nb_pt_r_tt"]].head()
pt_r_tt nb_pt_r_tt
0 139 8
1 133 8
2 133 8
3 144 8
4 121 7

Great, now we have those values in our accessibility GeoDataFrame. Let’s visualize the results and see how they look.

# Plot
accessibility_grid.plot(column="nb_pt_r_tt", linewidth=0, legend=True)
<AxesSubplot: >
../../_images/a12ede14c9293a38aca0402cb32646c7e183a145576757df5b5c80e841033a49.png

And here we go, now we have a map where we have used one of the common classifiers to classify our data into 9 classes.

Plotting a histogram#

A histogram is a graphic representation of the distribution of the data. When classifying the data, it’s always good to consider how the data is distributed, and how the classification shceme divides values into different ranges.

  • plot the histogram using pandas.DataFrame.plot.hist

  • Number of histogram bins (groups of data) can be controlled using the parameter bins:

# Histogram for public transport rush hour travel time
accessibility_grid["pt_r_tt"].plot.hist(bins=50)
<AxesSubplot: ylabel='Frequency'>
../../_images/b6a8324699a94e220a1207759b76de232b0ea365a7bff47f44f92c7f1a6d79f0.png

Let’s also add threshold values on thop of the histogram as vertical lines.

  • Natural Breaks:

import matplotlib.pyplot as plt

# Define classifier
classifier = mapclassify.NaturalBreaks(y=accessibility_grid["pt_r_tt"], k=9)

# Plot histogram for public transport rush hour travel time
accessibility_grid["pt_r_tt"].plot.hist(bins=50)

# Add vertical lines for class breaks
for break_point in classifier.bins:
    plt.axvline(break_point, color="k", linestyle="dashed", linewidth=1)
../../_images/5d529b4d7b0a9fcb065b7a910bdb20da19138f3595986c31bc97ca3d7501697c.png
  • Quantiles:

# Define classifier
classifier = mapclassify.Quantiles(y=accessibility_grid['pt_r_tt'])

# Plot histogram for public transport rush hour travel time
accessibility_grid["pt_r_tt"].plot.hist(bins=50)

for break_point in classifier.bins:
    plt.axvline(break_point, color="k", linestyle="dashed", linewidth=1)
../../_images/0dcfbe7f44fa3fdcef199daaba8b73d0247a69d9776974ac7f49eafe3eb6812c.png

Check your understanding

Select another column from the data (for example, travel times by car: car_r_t). Do the following visualizations using one of the classification schemes available from pysal/mapclassify:

  • histogram with vertical lines showing the classification bins

  • thematic map using the classification scheme

Applying a custom classifier#

Multicriteria data classification#

Let’s classify the geometries into two classes based on a given threshold -parameter. If the area of a polygon is lower than the threshold value (e.g., a certain distance), the output column will get a value 0, if it is larger, it will get a value 1. This kind of classification is often called a binary classification.

To classify each row of our GeoDataFrame we can iterate over each row or we can apply a function for each row. In our case, we will apply a lambda function for each row in our GeoDataFrame, which returns a value based on the conditions we provide.

Let’s do our classification based on two criteria: and find out grid cells

  1. Grid cells where the travel time is lower or equal to 20 minutes

  2. and they are further away than 4 km (4000 meters) from the city center.

Let’s first see how to classify a single row:

accessibility_grid.iloc[0]["pt_r_tt"] < 20 and accessibility_grid.iloc[0]["walk_d"] > 4000
False
int(accessibility_grid.iloc[11293]["pt_r_tt"] < 20 and accessibility_grid.iloc[11293]["walk_d"] > 4000)
1

Let’s now apply this to our GeoDataFrame and save it to a column called "suitable_area":

accessibility_grid["suitable_area"] = accessibility_grid.apply(lambda row: int(row["pt_r_tt"] < 20 and row["walk_d"] > 4000), axis=1)
accessibility_grid.head()
car_m_d car_m_t car_r_d car_r_t from_id pt_m_d pt_m_t pt_m_tt pt_r_d pt_r_t pt_r_tt to_id walk_d walk_t geometry nb_pt_r_tt suitable_area
0 32297 43 32260 48 5785640 32616 116 147 32616 108 139 5975375 32164 459 POLYGON ((382000.000 6697750.000, 381750.000 6... 8 0
1 32508 43 32471 49 5785641 32822 119 145 32822 111 133 5975375 29547 422 POLYGON ((382250.000 6697750.000, 382000.000 6... 8 0
2 30133 50 31872 56 5785642 32940 121 146 32940 113 133 5975375 29626 423 POLYGON ((382500.000 6697750.000, 382250.000 6... 8 0
3 32690 54 34429 60 5785643 33233 125 150 33233 117 144 5975375 29919 427 POLYGON ((382750.000 6697750.000, 382500.000 6... 8 0
4 31872 42 31834 48 5787544 32127 109 126 32127 101 121 5975375 31674 452 POLYGON ((381250.000 6697500.000, 381000.000 6... 7 0

Okey we have new values in suitable_area -column.

  • How many Polygons are suitable for us? Let’s find out by using a Pandas function called value_counts() that return the count of different values in our column.

# Get value counts
accessibility_grid["suitable_area"].value_counts()
0    13011
1        9
Name: suitable_area, dtype: int64

Okay, so there seems to be nine suitable locations for us where we can try to find an appartment to buy.

  • Let’s see where they are located:

# Plot
accessibility_grid.plot(column="suitable_area", linewidth=0)
<AxesSubplot: >
../../_images/fbf86783d34d53684846d03f6f3f65f34573b5a223c8b35b9828b27e20f8ffe3.png

A-haa, okay so we can see that suitable places for us with our criteria seem to be located in the eastern part from the city center. Actually, those locations are along the metro line which makes them good locations in terms of travel time to city center since metro is really fast travel mode.