Is 4 the Optimal Number of Clusters?

The last chart (displayed below) left the question about an optimal number of clusters unanswered. Seems like 4 is the 'local maximum', but the value 5 is not significantly lower than 4. We need to consider both cases.

Let's watch the scatter plot of average January vs July temperatures in the case of 4 clusters.

# Import the libraries

import pandas as pd

import matplotlib.pyplot as plt

import seaborn as sns

from sklearn.cluster import SpectralClustering

# Read the data

data = pd.read_csv('https://codefinity-content-media.s3.eu-west-1.amazonaws.com/138ab9ad-aa37-4310-873f-0f62abafb038/Cities+weather.csv', index_col = 0)

# Create the model

model = SpectralClustering(n_clusters = 4, affinity = 'nearest_neighbors')

# Fit the data and predict the labels

data['prediction'] = model.fit_predict(data.iloc[:,2:14])

# Visualize the results

sns.scatterplot(x = 'Jan', y = 'Jul', hue = 'prediction', data = data)

plt.show()


              123456789101112131415161718
            
# Import the libraries
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.cluster import SpectralClustering

# Read the data
data = pd.read_csv('https://codefinity-content-media.s3.eu-west-1.amazonaws.com/138ab9ad-aa37-4310-873f-0f62abafb038/Cities+weather.csv', index_col = 0)

# Create the model
model = SpectralClustering(n_clusters = 4, affinity = 'nearest_neighbors')

# Fit the data and predict the labels
data['prediction'] = model.fit_predict(data.iloc[:,2:14])

# Visualize the results
sns.scatterplot(x = 'Jan', y = 'Jul', hue = 'prediction', data = data)
plt.show()

The clustering seems logical, it splits the cities into different disjoint groups. But what if we build the same chart but for 5 clusters? That will be your task!

Compito

Swipe to start coding

Import SpectralClustering function from sklearn.cluster.
Create a SpectralClustering model with 5 clusters using the 'nearest_neighbors' affinity.
Fit the 3-14 columns of data to the model and predict the labels. Save the result within the 'prediction' column of data.
Build the seaborn scatter plot with average January (column 'Jan') vs July (column 'Jul') temperatures for each cluster (column 'prediction').

Soluzione

# Import the libraries

import pandas as pd

import matplotlib.pyplot as plt

import seaborn as sns

from sklearn.cluster import SpectralClustering

# Read the data

data = pd.read_csv('https://codefinity-content-media.s3.eu-west-1.amazonaws.com/138ab9ad-aa37-4310-873f-0f62abafb038/Cities+weather.csv', index_col = 0)

# Create the model

model = SpectralClustering(n_clusters = 5, affinity = 'nearest_neighbors')

# Fit the data and predict the labels

data['prediction'] = model.fit_predict(data.iloc[:,2:14])

# Visualize the results

sns.scatterplot(x = 'Jan', y = 'Jul', hue = 'prediction', data = data)

plt.show()

Tutto è chiaro?

Grazie per i tuoi commenti!

Sezione 4. Capitolo 5

# Import the libraries

import pandas as pd

import matplotlib.pyplot as plt

import seaborn as sns

___

# Read the data

data = pd.read_csv('https://codefinity-content-media.s3.eu-west-1.amazonaws.com/138ab9ad-aa37-4310-873f-0f62abafb038/Cities+weather.csv', index_col = 0)

# Create the model

model = ___(___, ___)

# Fit the data and predict the labels

data['prediction'] = ___.___(___.___[:,___])

# Visualize the results

___.___(x = '___', y = '___', ___ = 'prediction', data = data)

___.___()