Python for AI & ML - Day 16: Advanced Visualization with Seaborn

Goal: Master Seaborn, a high-level Python visualization library built on Matplotlib, to create sophisticated statistical plots. Learn to visualize distributions, relationships, and categorical data for deeper insights in AI/ML workflows.

---

1. What is Seaborn?

Seaborn simplifies complex data visualization by providing:

• Statistical Plotting: Built-in functions for distributions, correlations, and regressions.

• Aesthetic Defaults: Attractive themes and color palettes.

• Pandas Integration: Directly plot from DataFrames.

• Multi-Plot Grids: Create complex visualizations with minimal code.

Why Use Seaborn Over Matplotlib?

• Fewer lines of code for complex plots.

• Better handling of DataFrames and categorical data.

• Advanced tools for exploring relationships (e.g., heatmaps, pair plots).

---

2. Installation & Setup

pip install seaborn  

Import libraries:

import seaborn as sns  
import matplotlib.pyplot as plt  
import pandas as pd  

---

3. Key Features

1. Statistical Distributions

• Distributions: Histograms, KDE plots, rug plots.

• Regression: Visualize trends with confidence intervals.

• Categorical Data: Box plots, violin plots, bar plots.

2. Built-in Themes

Set aesthetics globally:

sns.set_theme(style="whitegrid", palette="pastel", font_scale=1.2)  

Themes: darkgrid, whitegrid, dark, white, ticks.

3. Color Palettes

Use predefined or custom palettes:

sns.color_palette("husl", 8)  # HUSL palette with 8 colors  
sns.set_palette("viridis")     # Set global palette  

---

4. Advanced Visualization Techniques

1. Distribution Plots

KDE (Kernel Density Estimate) Plot:

sns.kdeplot(data=df, x="Age", hue="Survived", fill=True)  
plt.title("Age Distribution by Survival")  

Joint Plot:

sns.jointplot(data=df, x="Age", y="Fare", kind="hex")  

2. Categorical Plots

Box Plot:

sns.boxplot(data=df, x="Pclass", y="Fare", hue="Survived")  

Violin Plot:

sns.violinplot(data=df, x="Pclass", y="Age", split=True, hue="Sex")  

Swarm Plot:

sns.swarmplot(data=df, x="Pclass", y="Age", hue="Sex", dodge=True)  

3. Matrix Plots

Heatmap:

corr = df.corr()  
sns.heatmap(corr, annot=True, cmap="coolwarm", vmin=-1, vmax=1)  

Cluster Map:

sns.clustermap(corr, cmap="viridis", standard_scale=1)  

4. Pair Plots

Visualize pairwise relationships:

sns.pairplot(df, hue="Survived", diag_kind="kde")  

5. FacetGrid

Multi-plot grids based on data subsets:

g = sns.FacetGrid(df, col="Pclass", row="Sex", hue="Survived")  
g.map(sns.scatterplot, "Age", "Fare").add_legend()  

---

5. Customizing Plots

Titles & Labels

plt.title("Custom Title", fontsize=14, fontweight="bold")  
plt.xlabel("X-Axis Label", fontsize=12)  
plt.ylabel("Y-Axis Label", fontsize=12)  

Annotations

Highlight specific data points:

ax = sns.barplot(data=df, x="Pclass", y="Fare")  
ax.bar_label(ax.containers[0], fmt="%.1f")  # Add values to bars  

Themes & Context

sns.set_context("paper")  # Options: "talk", "poster", "notebook"  

---

6. Real-World Use Cases in AI/ML

1. Feature Analysis:

   • Visualize feature distributions (sns.histplot).

   • Identify correlations (sns.heatmap).

2. Model Evaluation:

   • Plot ROC curves (sns.lineplot).

   • Compare metrics across classes (sns.barplot).

3. Clustering:

   • Visualize clusters with pair plots (sns.pairplot).

4. Time Series:

   • Plot trends with confidence intervals (sns.lineplot).

---

7. Best Practices

1. Choose the Right Plot:

   • Relationships: Scatter plots, line plots.

   • Distributions: Histograms, KDE plots.

   • Categorical Data: Box plots, bar plots.

2. Simplify: Avoid clutter (e.g., too many hues or categories).

3. Color Wisely: Use palettes that are colorblind-friendly (e.g., "colorblind").

---

8. Practice Exercise

1. Load the Titanic dataset (sns.load_dataset("titanic")).

2. Create a pair plot colored by survival.

3. Plot a heatmap of correlations between numeric features.

4. Use FacetGrid to visualize age distributions by class and gender.

Solution:

# 1. Load data  
titanic = sns.load_dataset("titanic")  

# 2. Pair plot  
sns.pairplot(titanic, hue="survived", vars=["age", "fare", "pclass"])  

# 3. Heatmap  
corr = titanic.corr(numeric_only=True)  
sns.heatmap(corr, annot=True, cmap="coolwarm")  

# 4. FacetGrid  
g = sns.FacetGrid(titanic, col="pclass", row="sex")  
g.map(sns.histplot, "age", kde=True).add_legend()  

---

Key Takeaways

• Seaborn simplifies statistical visualization with minimal code.

• Use FacetGrid and pair plots to explore multi-dimensional relationships.

• Customize themes and palettes for professional-quality visuals.