Modeling by Species

As observed in our initial visualization, penguins cluster by species. Instead of one global model, we can fit individual regressions for each species.

1. Species-Specific Training

We split the data and train three distinct models for Adelie, Chinstrap, and Gentoo penguins.

# Get the name of each species
species_list = penguins['species'].unique()
models = {}

# For each species, fit a separate linear regression model
for species in species_list:
    # Select the subset of data for the current species
    subset = penguins[penguins['species'] == species]
    X_s = subset[["bill_length_mm"]]
    y_s = subset["body_mass_g"]
    
    # Fit the model
    model_s = LinearRegression().fit(X_s, y_s)
    models[species] = model_s

    # Print the R^2 score for each species-specific model
    print(f"{species} R^2 Score: {model_s.score(X_s, y_s):.3f}")

Adelie R^2 Score: 0.296
Chinstrap R^2 Score: 0.264
Gentoo R^2 Score: 0.445

2. Visual Comparison

Visualizing separate lines highlights the different relationships between bill length and body mass across species.

# Visualize the regression lines for all species
sns.lmplot(data=penguins, x="bill_length_mm", y="body_mass_g", hue="species")
plt.title("Linear Regression per Species")

Text(0.5, 1.0, 'Linear Regression per Species')

By analyzing each separately, we capture the unique biological traits of each group.