Housing Price Predictor for Buenos Aires 🇦🇷

Part 3: Predicting Housing Price with Neighborhood

__author__ = "Donald Ghazi"
__email__ = "donald@donaldghazi.com"
__website__ = "donaldghazi.com"

This project builds on my Predicting Housing Price with Size and Predicting Housing Price with Location projects. And in the Predicting Housing Price with Location project, I created a model that used location — represented by latitude and longitude — to predict price. In this project, I'm going to use a different representation for location: neighborhood.

import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import warnings

from glob import glob
from category_encoders import OneHotEncoder
from sklearn.linear_model import LinearRegression, Ridge  # noqa F401
from sklearn.metrics import mean_absolute_error
from sklearn.pipeline import make_pipeline
from sklearn.utils.validation import check_is_fitted

warnings.simplefilter(action="ignore", category=FutureWarning)

GOALS

• Build a model to predict apartment price based on neighborhood
• Encode categorical data to train linear model
• Battle overfitting using regularization

Machine Learning Workflow

• Prepare Data
  • Import: wrangle function & for loop
  • Explore: Extract neighborhood from string
  • Split
• Build Model
  • Baseline
  • Iterate: One-Hot Encoding
  • Evaluate: Overfitting
• Communicate Results
  • Curse of Dimensionality
  • Regularization: Ridge Regression
  • Bar Charts for Coefficients

Prepare Data

Import

def wrangle(filepath):
    # Read CSV file
    df = pd.read_csv(filepath)

    # Subset data: Apartments in "Capital Federal", less than 400,000
    mask_ba = df["place_with_parent_names"].str.contains("Capital Federal")
    mask_apt = df["property_type"] == "apartment"
    mask_price = df["price_aprox_usd"] < 400_000
    df = df[mask_ba & mask_apt & mask_price]

    # Subset data: Remove outliers for "surface_covered_in_m2"
    low, high = df["surface_covered_in_m2"].quantile([0.1, 0.9])
    mask_area = df["surface_covered_in_m2"].between(low, high)
    df = df[mask_area]

    # Split "lat-lon" column
    df[["lat", "lon"]] = df["lat-lon"].str.split(",", expand=True).astype(float)
    df.drop(columns="lat-lon", inplace=True)

    return df

In the Predicting Housing Price with Size & Predicting Housing Price with Location projects, I used my wrangle function to import two CSV files as DataFrames. And since I'm working with multiple CSV files to import in this project, I will instead use glob where I can read several CSV files into a single DataFrame.

The first step is to gather the names of all the files I want to import. I can do this using pattern matching.

Collect File Names

# Use glob to create a list that contains the filenames for all the Buenos Aires real estate CSVs
files = glob("data/buenos-aires-real-estate-*.csv") # Assign this list to the variable name files
files

['data/buenos-aires-real-estate-3.csv',
 'data/buenos-aires-real-estate-2.csv',
 'data/buenos-aires-real-estate-1.csv',
 'data/buenos-aires-real-estate-5.csv',
 'data/buenos-aires-real-estate-4.csv']

len(files)

5

The next step is to read each of the CSVs in files into a DataFrame, and put all of those DataFrames into a list. A for loop is a good way to iterate through files.

Wrangle With for Loop

frames = [] # Create an empty list called frames
for file in files:
    print(file)

data/buenos-aires-real-estate-3.csv
data/buenos-aires-real-estate-2.csv
data/buenos-aires-real-estate-1.csv
data/buenos-aires-real-estate-5.csv
data/buenos-aires-real-estate-4.csv

As I iterate through this for loop, it goes through one by one (all of the file names above). And I want to read those files into a DataFrame.

frames = [] 
for file in files:
    df = wrangle(file)
    print(df.shape)

(1288, 17)
(1315, 17)
(1343, 17)
(1331, 17)
(1305, 17)

frames = [] 
for file in files:
    df = wrangle(file)
    frames.append(df)

len(frames)

5

type(frames[0])

pandas.core.frame.DataFrame

frames[0].head()

	operation	property_type	place_with_parent_names	price	currency	price_aprox_local_currency	price_aprox_usd	surface_total_in_m2	surface_covered_in_m2	price_usd_per_m2	price_per_m2	floor	rooms	expenses	properati_url	lat	lon
7	sell	apartment	|Argentina|Capital Federal|Villa General Mitre|	120000.0	USD	1819488.00	120000.0	NaN	55.0	NaN	2181.818182	NaN	2.0	NaN	http://villa-general-mitre.properati.com.ar/xx...	-34.616004	-58.470506
20	sell	apartment	|Argentina|Capital Federal|Palermo|	89000.0	USD	1349453.60	89000.0	NaN	37.0	NaN	2405.405405	7.0	2.0	NaN	http://palermo.properati.com.ar/ya5i_venta_dep...	-34.584712	-58.444927
21	sell	apartment	|Argentina|Capital Federal|Saavedra|	183495.0	USD	2782224.58	183495.0	92.0	57.0	1994.51087	3219.210526	NaN	2.0	NaN	http://saavedra.properati.com.ar/12izq_venta_d...	-34.554652	-58.493644
41	sell	apartment	|Argentina|Capital Federal|Villa del Parque|	95000.0	USD	1440428.00	95000.0	53.0	47.0	1792.45283	2021.276596	NaN	2.0	NaN	http://villa-del-parque.properati.com.ar/wy0n_...	-34.610581	-58.479625
43	sell	apartment	|Argentina|Capital Federal|Belgrano|	95000.0	USD	1440428.00	95000.0	0.0	35.0	NaN	2714.285714	NaN	1.0	NaN	http://belgrano.properati.com.ar/xw9a_venta_de...	-34.558227	-58.458357

The final step is to use pandas to combine all the DataFrames in frames.

Concatenate DataFrames

# Use pd.concat to concatenate the items in frames into a single DataFrame df
df = pd.concat(frames, ignore_index=True) # Set the ignore_index argument to True
df.head()

	operation	property_type	place_with_parent_names	price	currency	price_aprox_local_currency	price_aprox_usd	surface_total_in_m2	surface_covered_in_m2	price_usd_per_m2	price_per_m2	floor	rooms	expenses	properati_url	lat	lon
0	sell	apartment	|Argentina|Capital Federal|Villa General Mitre|	120000.0	USD	1819488.00	120000.0	NaN	55.0	NaN	2181.818182	NaN	2.0	NaN	http://villa-general-mitre.properati.com.ar/xx...	-34.616004	-58.470506
1	sell	apartment	|Argentina|Capital Federal|Palermo|	89000.0	USD	1349453.60	89000.0	NaN	37.0	NaN	2405.405405	7.0	2.0	NaN	http://palermo.properati.com.ar/ya5i_venta_dep...	-34.584712	-58.444927
2	sell	apartment	|Argentina|Capital Federal|Saavedra|	183495.0	USD	2782224.58	183495.0	92.0	57.0	1994.51087	3219.210526	NaN	2.0	NaN	http://saavedra.properati.com.ar/12izq_venta_d...	-34.554652	-58.493644
3	sell	apartment	|Argentina|Capital Federal|Villa del Parque|	95000.0	USD	1440428.00	95000.0	53.0	47.0	1792.45283	2021.276596	NaN	2.0	NaN	http://villa-del-parque.properati.com.ar/wy0n_...	-34.610581	-58.479625
4	sell	apartment	|Argentina|Capital Federal|Belgrano|	95000.0	USD	1440428.00	95000.0	0.0	35.0	NaN	2714.285714	NaN	1.0	NaN	http://belgrano.properati.com.ar/xw9a_venta_de...	-34.558227	-58.458357

df.shape

(6582, 17)

Explore

Looking through the output from the df.head() call above, there's a little bit more cleaning I need to do before I can work with the neighborhood information in this dataset.

Since I'm using a wrangle function, I only need to change the function to re-clean all of my CSV files. Hence why functions are so useful.

Create "Neighborhood" Feature

# Find the neighborhood for each property in the "place_with_parent_names" column
df["place_with_parent_names"].head()

0    |Argentina|Capital Federal|Villa General Mitre|
1                |Argentina|Capital Federal|Palermo|
2               |Argentina|Capital Federal|Saavedra|
3       |Argentina|Capital Federal|Villa del Parque|
4               |Argentina|Capital Federal|Belgrano|
Name: place_with_parent_names, dtype: object

df["place_with_parent_names"].str.split("|").head()

0    [, Argentina, Capital Federal, Villa General M...
1            [, Argentina, Capital Federal, Palermo, ]
2           [, Argentina, Capital Federal, Saavedra, ]
3    [, Argentina, Capital Federal, Villa del Parqu...
4           [, Argentina, Capital Federal, Belgrano, ]
Name: place_with_parent_names, dtype: object

df["place_with_parent_names"].str.split("|", expand=True).head()

	0	1	2	3	4
0		Argentina	Capital Federal	Villa General Mitre
1		Argentina	Capital Federal	Palermo
2		Argentina	Capital Federal	Saavedra
3		Argentina	Capital Federal	Villa del Parque
4		Argentina	Capital Federal	Belgrano

df["place_with_parent_names"].str.split("|", expand=True)[3].head()

0    Villa General Mitre
1                Palermo
2               Saavedra
3       Villa del Parque
4               Belgrano
Name: 3, dtype: object

# Modify the wrangle function to create a new feature "neighborhood"

def wrangle(filepath):
    # Read CSV file
    df = pd.read_csv(filepath)

    # Subset data: Apartments in "Capital Federal", less than 400,000
    mask_ba = df["place_with_parent_names"].str.contains("Capital Federal")
    mask_apt = df["property_type"] == "apartment"
    mask_price = df["price_aprox_usd"] < 400_000
    df = df[mask_ba & mask_apt & mask_price]

    # Subset data: Remove outliers for "surface_covered_in_m2"
    low, high = df["surface_covered_in_m2"].quantile([0.1, 0.9])
    mask_area = df["surface_covered_in_m2"].between(low, high)
    df = df[mask_area]

    # Split "lat-lon" column
    df[["lat", "lon"]] = df["lat-lon"].str.split(",", expand=True).astype(float)
    df.drop(columns="lat-lon", inplace=True)
    
    # Extract neighborhood
    df["neighborhood"] = df["place_with_parent_names"].str.split("|", expand=True)[3]
    df.drop(columns="place_with_parent_names", inplace=True)

    return df

frames = [] 
for file in files:
    df = wrangle(file)
    frames.append(df)

df = pd.concat(frames, ignore_index=True)
df.head()

	operation	property_type	price	currency	price_aprox_local_currency	price_aprox_usd	surface_total_in_m2	surface_covered_in_m2	price_usd_per_m2	price_per_m2	floor	rooms	expenses	properati_url	lat	lon	neighborhood
0	sell	apartment	120000.0	USD	1819488.00	120000.0	NaN	55.0	NaN	2181.818182	NaN	2.0	NaN	http://villa-general-mitre.properati.com.ar/xx...	-34.616004	-58.470506	Villa General Mitre
1	sell	apartment	89000.0	USD	1349453.60	89000.0	NaN	37.0	NaN	2405.405405	7.0	2.0	NaN	http://palermo.properati.com.ar/ya5i_venta_dep...	-34.584712	-58.444927	Palermo
2	sell	apartment	183495.0	USD	2782224.58	183495.0	92.0	57.0	1994.51087	3219.210526	NaN	2.0	NaN	http://saavedra.properati.com.ar/12izq_venta_d...	-34.554652	-58.493644	Saavedra
3	sell	apartment	95000.0	USD	1440428.00	95000.0	53.0	47.0	1792.45283	2021.276596	NaN	2.0	NaN	http://villa-del-parque.properati.com.ar/wy0n_...	-34.610581	-58.479625	Villa del Parque
4	sell	apartment	95000.0	USD	1440428.00	95000.0	0.0	35.0	NaN	2714.285714	NaN	1.0	NaN	http://belgrano.properati.com.ar/xw9a_venta_de...	-34.558227	-58.458357	Belgrano

Split

I can condense the whole process (splitting data) down to one task.

Split Data

# Create feature matrix X_train and target vector y_train
target = "price_aprox_usd" # target is "price_aprox_usd"
features = ["neighborhood"] # X_train should contain one feature: "neighborhood"
y_train = df[target]
X_train = df[features]

Build Model

I can condense the code I use to establish my baseline.

Baseline Model

# Calculate the baseline mean absolute error for my model
y_mean = y_train.mean()
y_pred_baseline = [y_mean] * len(y_train)
print("Mean apt price:", y_mean)
print("Baseline MAE:", mean_absolute_error(y_train, y_pred_baseline))

Mean apt price: 132383.83701458524
Baseline MAE: 44860.10834274133

The mean apartment price and baseline MAE should be similar but not identical to one in Predicting Housing Price with Locationproject. This is because numbers change since I'm working with more data.

Iterate

If I try to fit a LinearRegression predictor to my training data at this point, I'll get an error that says "ValueError: could not convert string to float".

When I fit a linear regression model, I'm asking scikit-learn to perform a mathematical operation. The problem is that my training set contains neighborhood information in a non-numerical form.

In order to create my model I need to encode that information so that it's represented numerically. There are lots of transformers that can do this and for this project, I'll be using one from the Category Encoders library, called a OneHotEncoder.

Transformer: OneHotEncoder

X_train.head()

	neighborhood
0	Villa General Mitre
1	Palermo
2	Saavedra
3	Villa del Parque
4	Belgrano

# If I try to fit a LinearRegression predictor to my training data at this point, I'll get an error
model = LinearRegression()
model.fit(X_train, y_train)

# Instantiate a OneHotEncoder named ohe
ohe = OneHotEncoder(use_cat_names=True)

# Fit
ohe.fit(X_train)

# Transform
XT_train= ohe.transform(X_train)
print(XT_train.shape)
XT_train.head()

(6582, 57)

	neighborhood_Villa General Mitre	neighborhood_Palermo	neighborhood_Saavedra	neighborhood_Villa del Parque	neighborhood_Belgrano	neighborhood_Coghlan	neighborhood_Chacarita	neighborhood_Boedo	neighborhood_Caballito	neighborhood_Monserrat	...	neighborhood_Versalles	neighborhood_Villa Lugano	neighborhood_Parque Avellaneda	neighborhood_Las Cañitas	neighborhood_Villa Soldati	neighborhood_Parque Chas	neighborhood_Velez Sarsfield	neighborhood_Pompeya	neighborhood_Villa Real	neighborhood_Catalinas
0	1	0	0	0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	0	0	0
1	0	1	0	0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	0	0	0
2	0	0	1	0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	0	0	0
3	0	0	0	1	0	0	0	0	0	0	...	0	0	0	0	0	0	0	0	0	0
4	0	0	0	0	1	0	0	0	0	0	...	0	0	0	0	0	0	0	0	0	0

5 rows × 57 columns

Build Model

# Create a pipeline named model that contains a OneHotEncoder transformer and a LinearRegression predictor
model = make_pipeline(
    OneHotEncoder(use_cat_names=True),
    LinearRegression()
)

model.fit(X_train, y_train) # Fit my model to the training data

Pipeline(steps=[('onehotencoder',
                 OneHotEncoder(cols=['neighborhood'], use_cat_names=True)),
                ('linearregression', LinearRegression())])

In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.

# Check my work
check_is_fitted(model[-1])

Above, I built a model with two transformers and a predictor.

Evaluate

Regardless of how I build my model, the evaluation step stays the same. I'll see how my model performs with the training set.

Training Mean Absolute Error

# Create a list of predictions for the observations in the feature matrix X_train
y_pred_training = model.predict(X_train) # Name this list y_pred_training

# Calculate the training mean absolute error for the predictions in y_pred_training as compared to the true targets in y_train
mae_training = mean_absolute_error(y_train, y_pred_training) 
print("Training MAE:", round(mae_training, 2))

Training MAE: 39337.76

model = make_pipeline(
    OneHotEncoder(use_cat_names=True),
    Ridge()
)

model.fit(X_train, y_train)

Pipeline(steps=[('onehotencoder',
                 OneHotEncoder(cols=['neighborhood'], use_cat_names=True)),
                ('ridge', Ridge())])

In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.

y_pred_training = model.predict(X_train)
mae_training = mean_absolute_error(y_train, y_pred_training)
print("Training MAE:", round(mae_training, 2))

Training MAE: 39350.22

# Import test data "buenos-aires-test-features.csv" into a DataFrame
X_test = pd.read_csv("data/buenos-aires-test-features.csv")[features]

# Generate a Series of predictions using my model
y_pred_test = pd.Series(model.predict(X_test))
y_pred_test.head()

0    246624.694624
1    161355.968734
2     98232.051308
3    110846.030377
4    127777.538197
dtype: float64

Communicate Results

If I write out the equation for my model, it'll be too big to fit on the screen. That's because, when I used the OneHotEncoder to encode the neighborhood data, I created a much wider DataFrame, and each column/feature has it's own coefficient in my model's equation.

This is important to keep in mind for two reasons. First, it means that this is a high-dimensional model. Instead of a 2D or 3D plot, I'd need a 58-dimensional plot to represent it, which is impossible. Second, it means that I'll need to extract and represent the information for my equation a little differently than before. I'll start by getting my intercept and coefficient.

Extract Coefficients

# Extract coefficients for my model
model = make_pipeline(
    OneHotEncoder(use_cat_names=True),
    LinearRegression()
)

model.fit(X_train, y_train)

Pipeline(steps=[('onehotencoder',
                 OneHotEncoder(cols=['neighborhood'], use_cat_names=True)),
                ('linearregression', LinearRegression())])

In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.

intercept = model.named_steps["linearregression"].intercept_
coefficients = model.named_steps["linearregression"].coef_
print("coefficients len:", len(coefficients))
print(coefficients[:5])  # First five coefficients

coefficients len: 57
[2.73010108e+16 2.73010108e+16 2.73010108e+16 2.73010108e+16
 2.73010108e+16]

I have the values of my coefficients. Now, I will need to find out which features they belong to. I'll need to get that information by going into the part of my pipeline that did the encoding.

Extract Feature Names

# Extract the feature names of my encoded data from the OneHotEncoder in my model
feature_names = model.named_steps["onehotencoder"].get_feature_names()
print("features len:", len(feature_names))
print(feature_names[:5])  # First five feature names

features len: 57
['neighborhood_Villa General Mitre', 'neighborhood_Palermo', 'neighborhood_Saavedra', 'neighborhood_Villa del Parque', 'neighborhood_Belgrano']

I have coefficients and feature names, and now I need to put them together. For that, I'll use a Series.

Create Feature Importance Series

# Create a pandas Series named feat_imp where the index is my features and the values are my coefficients
feat_imp = pd.Series(coefficients, index=feature_names)
feat_imp.head()

neighborhood_Villa General Mitre    2.730101e+16
neighborhood_Palermo                2.730101e+16
neighborhood_Saavedra               2.730101e+16
neighborhood_Villa del Parque       2.730101e+16
neighborhood_Belgrano               2.730101e+16
dtype: float64

Print Equation

# Print the equation that my model has determined for predicting apartment price based on longitude and latitude
print(f"price = {intercept.round(2)}")
for f, c in feat_imp.items():
    print(f"+ ({round(c, 2)} * {f})")

price = -2.730101076170245e+16
+ (2.7301010761829004e+16 * neighborhood_Villa General Mitre)
+ (2.7301010761867264e+16 * neighborhood_Palermo)
+ (2.7301010761835744e+16 * neighborhood_Saavedra)
+ (2.7301010761808184e+16 * neighborhood_Villa del Parque)
+ (2.7301010761868016e+16 * neighborhood_Belgrano)
+ (2.730101076183328e+16 * neighborhood_Coghlan)
+ (2.7301010761817804e+16 * neighborhood_Chacarita)
+ (2.7301010761814068e+16 * neighborhood_Boedo)
+ (2.730101076183022e+16 * neighborhood_Caballito)
+ (2.7301010761800324e+16 * neighborhood_Monserrat)
+ (2.7301010761812824e+16 * neighborhood_Centro / Microcentro)
+ (2.7301010761863868e+16 * neighborhood_Nuñez)
+ (2.7301010761809816e+16 * neighborhood_Balvanera)
+ (2.7301010761826668e+16 * neighborhood_San Telmo)
+ (2.730101076182386e+16 * neighborhood_Almagro)
+ (2.7301010761876784e+16 * neighborhood_Barrio Norte)
+ (2.7301010761817296e+16 * neighborhood_Monte Castro)
+ (2.73010107618942e+16 * neighborhood_Recoleta)
+ (2.7301010761799344e+16 * neighborhood_Mataderos)
+ (2.7301010761833696e+16 * neighborhood_Villa Urquiza)
+ (2.7301010761804844e+16 * neighborhood_Parque Patricios)
+ (2.7301010761812304e+16 * neighborhood_Flores)
+ (2.730101076179223e+16 * neighborhood_Boca)
+ (2.730101076180674e+16 * neighborhood_Floresta)
+ (2.730101076181007e+16 * neighborhood_San Cristobal)
+ (2.7301010761827252e+16 * neighborhood_Villa Crespo)
+ (2.7301010761824684e+16 * neighborhood_Villa Devoto)
+ (2.7301010761808724e+16 * neighborhood_Villa Ortuzar)
+ (2.730101076182092e+16 * neighborhood_Villa Luro)
+ (2.730101076180119e+16 * neighborhood_)
+ (2.730101076181775e+16 * neighborhood_Once)
+ (2.73010107618201e+16 * neighborhood_Agronomía)
+ (2.730101076184857e+16 * neighborhood_Retiro)
+ (2.7301010761812904e+16 * neighborhood_Congreso)
+ (2.7301010761809956e+16 * neighborhood_San Nicolás)
+ (2.7301010761816084e+16 * neighborhood_Barracas)
+ (2.730101076181272e+16 * neighborhood_Villa Pueyrredón)
+ (2.730101076185966e+16 * neighborhood_Colegiales)
+ (2.7301010761806828e+16 * neighborhood_Liniers)
+ (2.7301010761951852e+16 * neighborhood_Puerto Madero)
+ (2.7301010761812964e+16 * neighborhood_Parque Chacabuco)
+ (2.7301010761812636e+16 * neighborhood_Tribunales)
+ (2.7301010761813676e+16 * neighborhood_Paternal)
+ (2.7301010761814468e+16 * neighborhood_Parque Centenario)
+ (2.7301010761778236e+16 * neighborhood_Constitución)
+ (2.7301010761800224e+16 * neighborhood_Villa Santa Rita)
+ (2.7301010761825412e+16 * neighborhood_Abasto)
+ (2.730101076181496e+16 * neighborhood_Versalles)
+ (2.730101076177072e+16 * neighborhood_Villa Lugano)
+ (2.730101076178937e+16 * neighborhood_Parque Avellaneda)
+ (2.7301010761895564e+16 * neighborhood_Las Cañitas)
+ (2.730101076174681e+16 * neighborhood_Villa Soldati)
+ (2.7301010761786132e+16 * neighborhood_Parque Chas)
+ (2.7301010761786852e+16 * neighborhood_Velez Sarsfield)
+ (2.730101076176819e+16 * neighborhood_Pompeya)
+ (2.7301010761811052e+16 * neighborhood_Villa Real)
+ (2.730101076177681e+16 * neighborhood_Catalinas)

Curse of Dimensionality & Ridge

# Change the predictor in my model to Ridge and retrain it
model = make_pipeline(
    OneHotEncoder(use_cat_names=True),
    Ridge()
)
model.fit(X_train, y_train)

Pipeline(steps=[('onehotencoder',
                 OneHotEncoder(cols=['neighborhood'], use_cat_names=True)),
                ('ridge', Ridge())])

In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.

y_pred_training = model.predict(X_train)
mae_training = mean_absolute_error(y_train, y_pred_training)
print("Training MAE:", round(mae_training, 2))

Training MAE: 39350.22

X_test = pd.read_csv("data/buenos-aires-test-features.csv")[features]
y_pred_test = pd.Series(model.predict(X_test))
y_pred_test.head()

0    246624.694624
1    161355.968734
2     98232.051308
3    110846.030377
4    127777.538197
dtype: float64

intercept = model.named_steps["ridge"].intercept_
coefficients = model.named_steps["ridge"].coef_
print("coefficients len:", len(coefficients))
print(coefficients[:5])  # First five coefficients

coefficients len: 57
[  7714.62159938  45934.41058509  14701.16243685 -12595.50084744
  46954.20800904]

feature_names = model.named_steps["onehotencoder"].get_feature_names()
print("features len:", len(feature_names))
print(feature_names[:5])  # First five feature names

features len: 57
['neighborhood_Villa General Mitre', 'neighborhood_Palermo', 'neighborhood_Saavedra', 'neighborhood_Villa del Parque', 'neighborhood_Belgrano']

feat_imp = pd.Series(coefficients, index=feature_names)
feat_imp.head()

neighborhood_Villa General Mitre     7714.621599
neighborhood_Palermo                45934.410585
neighborhood_Saavedra               14701.162437
neighborhood_Villa del Parque      -12595.500847
neighborhood_Belgrano               46954.208009
dtype: float64

Feature Importance Bar Chart

feat_imp.head()

neighborhood_Villa General Mitre     7714.621599
neighborhood_Palermo                45934.410585
neighborhood_Saavedra               14701.162437
neighborhood_Villa del Parque      -12595.500847
neighborhood_Belgrano               46954.208009
dtype: float64

I'm back on track with my model, so now I can create a visualization that will help me understand what the most important features for my model in predicting apartment price.

feat_imp.sort_values()

neighborhood_Villa Soldati           -59248.813857
neighborhood_Villa Lugano            -48669.354978
neighborhood_Pompeya                 -43909.591101
neighborhood_Constitución            -41748.733031
neighborhood_Parque Chas             -32439.870833
neighborhood_Parque Avellaneda       -29585.606863
neighborhood_Boca                    -28353.364910
neighborhood_Velez Sarsfield         -27219.717857
neighborhood_Catalinas               -22012.323660
neighborhood_Mataderos               -21078.779668
neighborhood_Monserrat               -20292.596013
neighborhood_Villa Santa Rita        -19843.919642
neighborhood_                        -19370.743498
neighborhood_Parque Patricios        -15807.014824
neighborhood_Floresta                -14088.018047
neighborhood_Liniers                 -13729.104399
neighborhood_Villa del Parque        -12595.500847
neighborhood_Villa Ortuzar           -11208.901798
neighborhood_Balvanera               -11172.552799
neighborhood_San Nicolás             -10734.348396
neighborhood_San Cristobal           -10678.633615
neighborhood_Flores                   -8662.283049
neighborhood_Villa Pueyrredón         -8093.450148
neighborhood_Congreso                 -7974.663425
neighborhood_Centro / Microcentro     -7905.294912
neighborhood_Tribunales               -7818.086935
neighborhood_Parque Chacabuco         -7678.616944
neighborhood_Villa Real               -7393.485490
neighborhood_Paternal                 -7108.232790
neighborhood_Boedo                    -6837.403741
neighborhood_Parque Centenario        -6323.679527
neighborhood_Versalles                -4937.206101
neighborhood_Barracas                 -4618.655661
neighborhood_Monte Castro             -3427.438161
neighborhood_Once                     -3230.374461
neighborhood_Chacarita                -2898.959335
neighborhood_Agronomía                 -772.703682
neighborhood_Villa Luro                  -6.295553
neighborhood_Almagro                   2903.335566
neighborhood_Villa Devoto              3860.580177
neighborhood_Abasto                    4330.546318
neighborhood_San Telmo                 5638.469331
neighborhood_Villa Crespo              6277.048500
neighborhood_Villa General Mitre       7714.621599
neighborhood_Caballito                 9252.890876
neighborhood_Coghlan                  12223.107722
neighborhood_Villa Urquiza            12671.714625
neighborhood_Saavedra                 14701.162437
neighborhood_Retiro                   27042.606187
neighborhood_Colegiales               38436.327903
neighborhood_Nuñez                    42831.321413
neighborhood_Palermo                  45934.410585
neighborhood_Belgrano                 46954.208009
neighborhood_Barrio Norte             55590.928973
neighborhood_Las Cañitas              72270.212270
neighborhood_Recoleta                 72740.780756
neighborhood_Puerto Madero           128100.047303
dtype: float64

feat_imp.sort_values(key=abs)

neighborhood_Villa Luro                  -6.295553
neighborhood_Agronomía                 -772.703682
neighborhood_Chacarita                -2898.959335
neighborhood_Almagro                   2903.335566
neighborhood_Once                     -3230.374461
neighborhood_Monte Castro             -3427.438161
neighborhood_Villa Devoto              3860.580177
neighborhood_Abasto                    4330.546318
neighborhood_Barracas                 -4618.655661
neighborhood_Versalles                -4937.206101
neighborhood_San Telmo                 5638.469331
neighborhood_Villa Crespo              6277.048500
neighborhood_Parque Centenario        -6323.679527
neighborhood_Boedo                    -6837.403741
neighborhood_Paternal                 -7108.232790
neighborhood_Villa Real               -7393.485490
neighborhood_Parque Chacabuco         -7678.616944
neighborhood_Villa General Mitre       7714.621599
neighborhood_Tribunales               -7818.086935
neighborhood_Centro / Microcentro     -7905.294912
neighborhood_Congreso                 -7974.663425
neighborhood_Villa Pueyrredón         -8093.450148
neighborhood_Flores                   -8662.283049
neighborhood_Caballito                 9252.890876
neighborhood_San Cristobal           -10678.633615
neighborhood_San Nicolás             -10734.348396
neighborhood_Balvanera               -11172.552799
neighborhood_Villa Ortuzar           -11208.901798
neighborhood_Coghlan                  12223.107722
neighborhood_Villa del Parque        -12595.500847
neighborhood_Villa Urquiza            12671.714625
neighborhood_Liniers                 -13729.104399
neighborhood_Floresta                -14088.018047
neighborhood_Saavedra                 14701.162437
neighborhood_Parque Patricios        -15807.014824
neighborhood_                        -19370.743498
neighborhood_Villa Santa Rita        -19843.919642
neighborhood_Monserrat               -20292.596013
neighborhood_Mataderos               -21078.779668
neighborhood_Catalinas               -22012.323660
neighborhood_Retiro                   27042.606187
neighborhood_Velez Sarsfield         -27219.717857
neighborhood_Boca                    -28353.364910
neighborhood_Parque Avellaneda       -29585.606863
neighborhood_Parque Chas             -32439.870833
neighborhood_Colegiales               38436.327903
neighborhood_Constitución            -41748.733031
neighborhood_Nuñez                    42831.321413
neighborhood_Pompeya                 -43909.591101
neighborhood_Palermo                  45934.410585
neighborhood_Belgrano                 46954.208009
neighborhood_Villa Lugano            -48669.354978
neighborhood_Barrio Norte             55590.928973
neighborhood_Villa Soldati           -59248.813857
neighborhood_Las Cañitas              72270.212270
neighborhood_Recoleta                 72740.780756
neighborhood_Puerto Madero           128100.047303
dtype: float64

# Create a horizontal bar chart that shows the top 15 coefficients for my model, based on their absolute value
feat_imp.sort_values(key=abs).tail(15).plot(kind="barh")
plt.xlabel("Importance [USD]")
plt.ylabel("Feature")
plt.title("Feature Importance for Apartment Price");

Looking at this bar chart, I can see that the poshest neighborhoods in Buenos Aires like Puerto Madero and Recoleta increase the predicted price of an apartment, while more working-class neighborhoods like Villa Soldati and Villa Lugano decrease the predicted price.