Bivariate Analysis in EDA: Best Plots to Understand Relationships Between Variables

After understanding individual features using univariate analysis, the next important step in Exploratory Data Analysis (EDA) is Bivariate Analysis.

Bivariate analysis focuses on analyzing the relationship between two variables. This step is crucial because machine learning models rely heavily on how features interact with each other and with the target variable.

If you skip this step, you may miss important patterns, correlations, and dependencies that directly affect your model’s performance.

What is Bivariate Analysis?

Bivariate Analysis means studying the relationship between two variables.

Examples:

• Age vs Salary
• Experience vs Salary
• Study Hours vs Marks
• Product Category vs Sales

The goal is to understand:

• Whether variables are related
• The strength of the relationship
• The direction (positive or negative)
• Patterns or trends
• Differences between categories
• 
  

Why Bivariate Analysis is Important

Many beginners only focus on individual features, but real insights come from relationships.

Benefits:

• Helps in feature selection
• Improves model accuracy
• Identifies strong predictors
• Detects hidden patterns
• Helps avoid multicollinearity
• 
  

Example:

If experience and salary are strongly correlated, your model can use that information effectively.

Best Plots for Bivariate Analysis

Let’s explore the most useful plots and when to use them.

1. Scatter Plot

Best For:

Numerical vs Numerical

Use When:

You want to see the relationship between two continuous variables.

What It Shows:

• Positive or negative correlation
• Clusters
• Outliers

Code:

sns.scatterplot(x="Age", y="Salary", data=df)

2. Line Plot

Best For:

Trends over time or ordered data

Use When:

You want to track changes between two variables over time.

Code:

sns.lineplot(x="Year", y="Sales", data=df)

3. Bar Plot

Best For:

Categorical vs Numerical

Use When:

You want to compare average values across categories.

Code:

sns.barplot(x="Category", y="Sales", data=df)

4. Boxplot

Best For:

Categorical vs Numerical

Use When:

You want to compare distributions across categories.

What It Shows:

• Spread of data
• Median
• Outliers

Code:

sns.boxplot(x="Category", y="Price", data=df)

5. Violin Plot

Best For:

Distribution comparison

Use When:

You want more detailed distribution insights than boxplot.

Code:

sns.violinplot(x="Category", y="Score", data=df)

6. Strip Plot

Best For:

Showing individual data points

Use When:

You want to visualize distribution clearly.

Code:

sns.stripplot(x="Category", y="Value", data=df)

7. Swarm Plot

Best For:

Better version of strip plot

Use When:

You want to avoid overlapping points.

Code:

sns.swarmplot(x="Category", y="Value", data=df)

8. Heatmap

Best For:

Correlation analysis

Use When:

You want to visualize relationships between multiple numerical variables.

What It Shows:

• Correlation strength
• Positive/negative relationships

Code:

sns.heatmap(df.corr(), annot=True)

9. Regression Plot

Best For:

Trend + relationship

Use When:

You want to see correlation along with a trend line.

Code:

sns.regplot(x="Age", y="Salary", data=df)

How to Choose the Right Plot

Numerical vs Numerical:

• Scatter Plot
• Line Plot
• Regression Plot

Categorical vs Numerical:

• Bar Plot
• Boxplot
• Violin Plot

Multiple Relationships:

• Heatmap

Distribution with categories:

• Strip Plot
• Swarm Plot

Real Example in Machine Learning

Suppose you are building a sales prediction model.

You can use:

• Scatter plot → Sales vs Advertising
• Bar plot → Sales by category
• Heatmap → Correlation between features
• Regression plot → Trend analysis

This helps you select better features and build a stronger model.

Common Mistakes Beginners Make

1. Ignoring Relationships

Only focusing on single variables.

2. Using Wrong Plot

Using bar plot instead of scatter for numerical data.

3. Not Checking Correlation

Missing important relationships.

4. Overlooking Outliers

Outliers can distort relationships.

Final Thoughts

Bivariate analysis is where real insights begin. It helps you understand how variables interact, which is essential for building accurate machine learning models.

Before moving to modeling, always analyze relationships properly.

Remember:

Good EDA = Better Insights = Better Models

Mastering bivariate analysis will take your data science skills to the next level.