Data Visualization

Problem

When structured data needs to become a chart, the hard part is often not rendering, but choosing the right visual form from variable types, the analytical question, and the output medium. Feeding raw data directly into matplotlib can produce unreadable axis labels, inconsistent colors, missing-value gaps, or bar charts that exaggerate differences.

How It Works

The skill breaks visualization into explicit steps: inspect whether fields are numeric, categorical, or temporal, then select chart types based on comparison, trend, relationship, distribution, correlation, or composition. It then aggregates, pivots, sorts, resamples, and removes NaN values before drawing. Styling focuses on readable axis labels, units, legends, removal of unnecessary gridlines and borders, and size choices matched to reports or dashboards. Static output uses matplotlib and seaborn, saved as PNG or SVG; interactive output uses plotly with hover, zoom, and pan.

Boundaries and Notes

It is suited for analytics charts in reports and documentation, not for statistical modeling or full dashboard frontend development. Large category sets may need an Other bucket or a treemap; dense scatter plots may need alpha, jitter, or hexbin/density views; missing values, long labels, and skewed data each need specific handling. If no chart type is specified, it infers one from the question, but providing audience and output format is still recommended.