Histogram

A histogram is one of the quickest ways to “see” what your data looks like. Instead of staring at a long list of numbers, you get a picture of how values are spread out and where they tend to cluster.

What it is
A histogram is a bar chart for a single numeric variable (like house prices or exam scores). You split the number line into intervals called bins (for example, 0–10, 10–20, 20–30), then count how many observations fall into each bin. The height (or area) of each bar represents that frequency (or sometimes density), and the bars touch because the bins represent continuous ranges.

How to read it
A histogram helps you notice distribution features that matter in practice:

• Center: where most values sit (a “typical” range).
• Spread: whether values are tightly packed or widely scattered.
• Skew: a long tail to the right (few very large values) or left.
• Modes: one peak vs. multiple peaks (possible subgroups).
• Outliers: rare extreme values that may need investigation.

Why bin choice matters
The bin width can change the story: very wide bins can hide structure, while very narrow bins can make random noise look like patterns. Common automatic choices include Sturges’ rule or the Freedman–Diaconis rule, but it’s often worth trying a couple of bin sizes.

Why it matters in AI/ML
Histograms are a go-to step in exploratory data analysis (EDA). They reveal skewed features that may benefit from log transforms, highlight data leakage or weird measurement artifacts, and show whether a target variable (like sales) has heavy tails that can affect loss functions and model stability. They’re also used directly in methods like Histogram of Oriented Gradients (HOG) and in monitoring feature drift in production.