Visualization Cheat Sheet¶

matplotlib and seaborn essentials. The full explanation is in Data visualization.

Setup and the object model¶

import matplotlib.pyplot as plt
import seaborn as sns

sns.set_theme(style="whitegrid", palette="colorblind")   # optional global style

fig, ax = plt.subplots(figsize=(7, 4))                      # one plot
fig, axes = plt.subplots(2, 3, figsize=(12, 6),             # a grid: axes[row, col]
                         sharex=True, sharey=True)
fig.tight_layout()                                          # fix overlapping labels
fig.savefig("chart.png", dpi=150, bbox_inches="tight")     # .svg/.pdf for vector
plt.show()

pyplot (implicit)	Object-oriented (explicit)
`plt.plot(x, y)`	`ax.plot(x, y)`
`plt.title("t")`	`ax.set_title("t")`
`plt.xlabel("x")`	`ax.set_xlabel("x")`
`plt.xlim(0, 10)`	`ax.set_xlim(0, 10)`
`plt.xticks(...)`	`ax.set_xticks(...)`
`plt.legend()`	`ax.legend()`

Core matplotlib plots¶

Plot	Call	Key options
Line	`ax.plot(x, y)`	`lw`, `ls="--"`, `marker="o"`, `label`
Scatter	`ax.scatter(x, y)`	`s` (size), `c` (color values), `alpha`, `cmap`
Bar	`ax.bar(cats, vals)` / `ax.barh(...)`	`width`, `color`, `yerr`
Histogram	`ax.hist(x, bins=30)`	`density=True`, `alpha`, `histtype="step"`
Box	`ax.boxplot([a, b])`	`labels`, `showfliers`
Image / matrix	`ax.imshow(M)`	`cmap`, `vmin`, `vmax`, `aspect`
Fill band	`ax.fill_between(x, lo, hi)`	`alpha` (for confidence bands)
Reference line	`ax.axhline(y)` / `ax.axvline(x)`	`color`, `ls`
Shaded span	`ax.axvspan(x0, x1)`	`alpha`
Text	`ax.text(x, y, "s")` / `ax.annotate("s", xy, xytext)`	`ha`, `va`, `arrowprops`

Styling details¶

ax.set_title("Takeaway goes here", loc="left", fontweight="bold")
ax.set(xlabel="Distance (km)", ylabel="Time (min)", ylim=(0, None))
ax.spines[["top", "right"]].set_visible(False)       # remove chart junk
ax.set_xscale("log")                                  # log axis
ax.tick_params(axis="x", rotation=45)
ax.legend(frameon=False, loc="upper left")
ax.grid(axis="y", alpha=0.3)
ax.yaxis.set_major_formatter("{x:,.0f}")              # thousands separators
twin = ax.twinx()                                     # second y-axis (use sparingly)

seaborn: axes-level functions¶

These draw on one Axes. Pass ax= to choose which.

Purpose	Function	Example
Histogram / KDE	`histplot`, `kdeplot`	`sns.histplot(df, x="v", hue="g", kde=True, stat="density")`
Scatter	`scatterplot`	`sns.scatterplot(df, x="a", y="b", hue="g", size="n", alpha=.6)`
Line with CI	`lineplot`	`sns.lineplot(df, x="t", y="v", hue="g", errorbar=("ci", 95))`
Category means	`barplot`, `pointplot`	`sns.barplot(df, x="g", y="v", errorbar=("ci", 95))`
Counts	`countplot`	`sns.countplot(df, y="cat", order=df["cat"].value_counts().index)`
Distributions by group	`boxplot`, `violinplot`, `stripplot`	`sns.boxplot(df, x="g", y="v")`
Matrix	`heatmap`	`sns.heatmap(corr, annot=True, fmt=".2f", cmap="vlag", center=0)`
Regression fit	`regplot`	`sns.regplot(df, x="a", y="b", ci=None)`

seaborn: figure-level functions (small multiples)¶

These create their own figure and return a FacetGrid. Use col=, row=, and col_wrap=.

Function	Wraps	Example
`relplot`	scatter, line	`sns.relplot(df, x="a", y="b", col="g", kind="scatter")`
`displot`	hist, kde, ecdf	`sns.displot(df, x="v", col="g", row="h", bins=20)`
`catplot`	box, bar, violin, strip	`sns.catplot(df, x="g", y="v", col="h", kind="box")`
`lmplot`	regplot	`sns.lmplot(df, x="a", y="b", col="g", col_wrap=3)`
`pairplot`	all pairs	`sns.pairplot(df[cols + ["target"]], hue="target")`

g = sns.relplot(df, x="a", y="b", col="g", height=3, aspect=1.2)
g.set_titles("{col_name}")
g.set_axis_labels("A (units)", "B (units)")
g.figure.suptitle("Title", y=1.03)

Which chart for which question¶

Question	Chart
Change over time	line; add a rolling mean for noisy series
Compare categories	bar (horizontal for long labels), sorted, starting at 0
One distribution	histogram (try several bin counts), KDE, ECDF
Distributions across groups	box, violin, overlaid density, small multiples
Relationship of two numbers	scatter (`alpha` for overplotting, `hexbin` for very many points)
Many pairwise relationships	pair plot, correlation heatmap
Parts of a whole	stacked bar, or a sorted bar of shares; pie only for 2–3 parts
Model results	predicted vs actual scatter, residual plot, confusion matrix heatmap

Color¶

Data type	Colormap kind	Good choices
Ordered values	sequential	`viridis`, `cividis`, `rocket`, `Blues`
Values around a midpoint	diverging	`vlag`, `RdBu`, `coolwarm`, with `center=`
Categories	qualitative	`sns.color_palette("colorblind")`, `tab10`

palette = sns.color_palette("colorblind", 3)
sns.palplot(palette)                        # preview a palette
ax.scatter(x, y, c=values, cmap="viridis"); fig.colorbar(ax.collections[0], ax=ax)

Avoid jet and rainbow maps, and don't rely on red versus green alone. Add markers, line styles, or direct labels as a second cue.

pandas plotting shortcuts¶

df["col"].hist(bins=30)
df.hist(bins=30, figsize=(12, 8))                     # every numeric column
df["cat"].value_counts().plot.barh()
df.plot(x="date", y=["a", "b"], figsize=(10, 4))      # lines
df.plot.scatter(x="a", y="b", alpha=0.5)
df.boxplot(column="v", by="g")

All of these return a matplotlib Axes, so you can keep customizing.

Plotly (interactive)¶

import plotly.express as px

fig = px.scatter(df, x="a", y="b", color="g", hover_data=["id"], log_x=True)
fig = px.line(df, x="date", y="v", color="g")
fig = px.histogram(df, x="v", color="g", barmode="overlay", nbins=40)
fig.update_layout(title="Title", template="plotly_white")
fig.show()
fig.write_html("chart.html")

Explanatory chart checklist¶

The title states the takeaway, not just the variables.
Axes are labeled, with units.
Bars start at zero.
Categories are sorted (unless they have a natural order).
Lines are labeled directly instead of with a legend, where possible.
The palette is colorblind-safe, with a second cue (markers or line styles).
Clutter is removed: top and right spines, heavy grids, 3D effects.
Text is large enough to read when the image is shrunk into a slide.