Section 1.1: What is a Correlation Matrix?

A correlation matrix serves as a visual tool that illustrates the relationships between various variables. It displays the strength of these relationships in a grid format, indicating whether they are positive or negative. This matrix is invaluable for identifying trends in your dataset and detecting anomalies.

Section 1.2: The Dataset

For our analysis, we'll utilize a dataset featuring the menu items from McDonald's India. You can download it from Kaggle to follow along with our process:

McDonald’s India: Menu Nutrition Dataset | Kaggle

Let’s load the dataset and explore its structure:

Image by the author: excerpt from the dataset.

Section 1.3: Plotting the Correlation Matrix with Seaborn

We will leverage the correlation matrix to visualize the relationships among the various features within our dataset. Initially, we will compute the Pearson correlation, which measures the strength of a linear relationship between two variables. The values range from -1 (indicating a total negative correlation) to +1 (indicating a total positive correlation), with 0 signifying no correlation. We will round these values to two decimal places using the Pandas library's .corr method.

Now, let’s visualize this using the Seaborn library.

To create the plot, we can set the figure size and DPI (dots per inch) for high resolution. We utilize the sns.heatmap method for the correlation matrix and define parameters such as:

• square: Ensures each cell is square-shaped.
• vmin, vmax, center: Sets the value range for the colormap.
• cmap: Chooses the 'RdBu_r' diverging colormap.
• cbar_kws: Adjusts the size of the colorbar.

Analysis of the results indicates strong positive correlations among energy, protein, sodium, total fat, and saturated fat, which may suggest a less healthy dietary profile. Additionally, there is a notable correlation between energy and total carbohydrate levels, as well as between total sugar and added sugar content.

Section 2.1: First Variant with Heatmapz

An exciting library called Heatmapz can be installed via pip. This library allows us to incorporate size as a parameter in our heatmap, where the size of each square reflects the strength of the correlation it represents.

This adjustment makes it easier to focus on areas of significant correlation, as weaker correlations become less prominent. The size of the squares visually indicates the magnitude of the correlation.

Section 2.2: Visualizing Correlation with Bar Charts

While bar charts may not always be the most visually appealing, they effectively convey information. Let’s explore creating a bar chart for correlation visualization. We will develop two functions: one to remove duplicates and another to generate the desired pairs for visualization.

Here’s the result we achieve:

Image by the author.

Let’s plot the bar chart now!

Section 2.3: Conclusion

Choose the correlation matrix visualization style that best suits your needs. Personally, I find the bar chart to be the most straightforward and easy to interpret. I encourage you to experiment with these various techniques and discover what works best for your data analysis.

For more insights, subscribe to my free weekly newsletter, “The Mindful Data Path,” where I simplify finance topics and guide you in practicing Data Science with me.

References:

heatmapz · PyPI

seaborn.heatmap — seaborn 0.11.2 documentation (pydata.org)

python — List Highest Correlation Pairs from a Large Correlation Matrix in Pandas? — Stack Overflow

matplotlib.colors.TwoSlopeNorm — Matplotlib 3.5.2 documentation

Throw out the correlation matrix and use bar charts to visualize correlation? ? | Levi (typefully.com)

More content at PlainEnglish.io. Sign up for our free weekly newsletter. Follow us on Twitter, LinkedIn, YouTube, and Discord.