Project Description

In this project, we will explore data about penguins in Antarctica using unsupervised learning. Our goal is to discover hidden patterns and natural groupings (clusters) among the penguins based on their physical features.

---

Background

A team of researchers studying penguins in Antarctica has been collecting data, and they’ve asked for our help! They believe there are at least three penguin species—Adelie, Chinstrap, and Gentoo—but unfortunately, they did not record the species labels in the dataset.

We’ve been given the dataset in CSV format as:
penguins.csv

---

Dataset Information

Source:
Data collected by Dr. Kristen Gorman and the Palmer Station, Antarctica LTER, part of the Long Term Ecological Research Network.

Features Available:

Column	Description
culmen_length_mm	Length of the penguin's bill (mm)
culmen_depth_mm	Depth of the penguin's bill (mm)
flipper_length_mm	Length of the flipper (mm)
body_mass_g	Body mass (grams)
sex	Penguin’s sex (Male/Female)

---

Our Task

We will:

1. Explore and process the data.
2. Perform clustering analysis to identify natural groups of penguins.
3. Decide on a reasonable number of clusters (hint: we expect ~3 species).
4. Analyze and compare the average feature values for each cluster.

---

Final Outcome

By using unsupervised learning techniques like K-Means Clustering, we aim to:

• Group penguins into distinct clusters.
• Provide meaningful insights to the research team.
• Possibly help them identify penguin species based on the cluster patterns.

Let’s dive in!

# Import Required libraries
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.cluster import KMeans
from sklearn.preprocessing import StandardScaler

# Loading and examining the dataset
penguins_df = pd.read_csv("penguins.csv")
penguins_df.head()

	culmen_length_mm	culmen_depth_mm	flipper_length_mm	body_mass_g	sex
0	39.1	18.7	181.0	3750.0	MALE
1	39.5	17.4	186.0	3800.0	FEMALE
2	40.3	18.0	195.0	3250.0	FEMALE
3	36.7	19.3	193.0	3450.0	FEMALE
4	39.3	20.6	190.0	3650.0	MALE

penguins_df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 332 entries, 0 to 331
Data columns (total 5 columns):
 #   Column             Non-Null Count  Dtype  
---  ------             --------------  -----  
 0   culmen_length_mm   332 non-null    float64
 1   culmen_depth_mm    332 non-null    float64
 2   flipper_length_mm  332 non-null    float64
 3   body_mass_g        332 non-null    float64
 4   sex                332 non-null    object 
dtypes: float64(4), object(1)
memory usage: 13.1+ KB

# Convert categorical variables into dummy/indicator variables
penguins_df = pd.get_dummies(penguins_df, dtype='int')

# Scaling variables  
scaler = StandardScaler()
X = scaler.fit_transform(penguins_df)
penguins_preprocessed = pd.DataFrame(data=X,columns=penguins_df.columns)
penguins_preprocessed.head()

	culmen_length_mm	culmen_depth_mm	flipper_length_mm	body_mass_g	sex_FEMALE	sex_MALE
0	-0.903906	0.790360	-1.425342	-0.566948	-0.993994	0.993994
1	-0.830434	0.126187	-1.068577	-0.504847	1.006042	-1.006042
2	-0.683490	0.432728	-0.426399	-1.187953	1.006042	-1.006042
3	-1.344738	1.096901	-0.569105	-0.939551	1.006042	-1.006042
4	-0.867170	1.761074	-0.783164	-0.691149	-0.993994	0.993994

# Detect the optimal number of clusters for k-means clustering using elbow method
inertia = []
for k in range(1, 10):
    kmeans = KMeans(n_clusters=k, random_state=42).fit(penguins_preprocessed)
    inertia.append(kmeans.inertia_)    
plt.plot(range(1, 10), inertia, marker='o')
plt.xlabel('Number of clusters')
plt.ylabel('Inertia')
plt.title('Elbow Method')
plt.show()

n_clusters=4

# Run the k-means clustering with k=4 
kmeans = KMeans(n_clusters=n_clusters, random_state=42).fit(penguins_preprocessed)
penguins_df['label'] = kmeans.labels_

# visualize the clusters (here for the 'culmen_length_mm' column)
plt.scatter(penguins_df['label'], penguins_df['culmen_length_mm'], c=kmeans.labels_, cmap='viridis')
plt.xlabel('Cluster')
plt.ylabel('culmen_length_mm')
plt.xticks(range(int(penguins_df['label'].min()), int(penguins_df['label'].max()) + 1))
plt.title(f'K-means Clustering (K={n_clusters})')
plt.show()

# create `stat_penguins` DataFrame
numeric_columns = ['culmen_length_mm', 'culmen_depth_mm', 'flipper_length_mm','label']
stat_penguins = penguins_df[numeric_columns].groupby('label').mean()
stat_penguins

	culmen_length_mm	culmen_depth_mm	flipper_length_mm
label
0	43.878302	19.111321	194.764151
1	45.563793	14.237931	212.706897
2	40.217757	17.611215	189.046729
3	49.473770	15.718033	221.540984

Final Thoughts:

The clusters seem to be well-separated, meaning K-means has identified distinct groups. However, clusters 0, 1, and 2 have overlapping culmen lengths, which may indicate that K=4 is slightly high, or that another clustering method (e.g., DBSCAN or hierarchical clustering) might yield better results.

Cluster 3 (Yellow):

This cluster appears to have the highest culmen lengths, meaning these data points share a distinct trait that differentiates them from the rest.