Discretization
Discretization is the process of mapping values or observations into a finite set of discrete intervals or categories.
Financial AI in Practice
Binning is the technique of grouping continuous or categorical values into discrete intervals or categories.
- How it is used
- Makes credit-score inputs more interpretable and less sensitive to noise and outliers while supporting auditable lending decisions.
- How deep it goes
- Explains granularity trade-offs and special-value handling, then demonstrates manual quantile binning and constrained automated selection.
- Examples
- Uses age intervals, ten-bin `pd.qcut` numeric features, and a chart of bin counts and default rates.
- What you can do after
- Apply it via pandas `pd.qcut` or OptBinning `BinningProcess`, then inspect event rates, special bins, and selected features.
Outlier Detection in Python
Binning is the process of assigning continuous or otherwise detailed values to discrete intervals or categories.
- How it is used
- Prepares numeric or date values for outlier detectors that require categorical inputs or use histograms.
- How deep it goes
- Compares equal-width and equal-count methods, their uses, and sensitivity to extreme values and bin count.
- Examples
- Uses staff expense dates, times, and amounts, plus uniform data with outliers, to demonstrate binning choices.
- What you can do after
- Apply equal-width or equal-count bins while accounting for the detection task and possible effects of extreme values.
Read how Outlier Detection in Python explains discretization →
Grokking Machine Learning
Binning is a technique that partitions numerical values into intervals or buckets so they can be represented as categorical groups.
- How it is used
- Groups numerical features into categories when they may be more useful to a model as categorical features.
- How deep it goes
- Defines binning as splitting values into buckets and demonstrates Pandas interval creation, without evaluating bin choices.
- Examples
- Uses Titanic passenger ages and decade-wide intervals from 0 to 80 as the concrete example.
- What you can do after
- Apply it via Pandas `cut` to replace passenger ages with decade-based categorical intervals.
Read how Grokking Machine Learning explains discretization →
Machine Learning for Tabular Data: XGBoost, Deep Learning, and AI
A technique that groups numeric values into discrete ordered bins and assigns each value to a bin.
- How it is used
- Accelerates decision-tree and gradient-boosting split search, and supports categorical transformations for exploratory association analysis.
- How deep it goes
- Explains ordered integer bins, threshold evaluation, Gini scoring, and histogram speedups without developing the complete XGBoost split algorithm.
- Examples
- Uses latitude thresholds, Airbnb price deciles, MPG deciles, and a 256-bin histogram implementation.
- What you can do after
- Apply it via NumPy histogram operations to reduce split candidates or discretize numeric variables for classification and association analysis.
Read how Machine Learning for Tabular Data: XGBoost, Deep Learning, and AI explains discretization →
Grokking Statistics
Binning is the process of grouping continuous or discrete observations into intervals and summarizing each interval.
- How it is used
- Creates interval-level counts when raw records do not provide a usable dependent variable for correlation or regression.
- How deep it goes
- Demonstrates interval construction and counting, then stops short of endorsing linear analysis when outliers violate assumptions.
- Examples
- Uses airplane bird-strike records grouped into 100-foot altitude intervals and plotted against incident counts.
- What you can do after
- Apply it via pandas to create intervals, count observations, visualize the result, and question subsequent model choices.
Related concepts
- equal-width binningincludes · Equal-width binning is a discretization technique that partitions a numerical range into intervals of identical width.
- preprocessingkind of · Preprocessing is the preparation or transformation of inputs, data, content, or a model before a subsequent computational operation.
- continuous datauses · Continuous data consists of numerical measurements that can take arbitrarily fine-grained values across a range or interval.
- Chronosused by · Chronos is a foundation model and framework for probabilistic time-series forecasting that adapts language-model methods to temporal data.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.