Feature engineering
Feature engineering is the creation, transformation, or selection of input features from data to provide useful representations for analysis or modeling.
Machine Learning for Tabular Data: XGBoost, Deep Learning, and AI
Feature processing is the transformation or elaboration of input features to make them suitable or more informative for machine learning models.
- How it is used
- Prepares tabular inputs for predictive modeling by extracting relevant information and aligning representations with model needs.
- How deep it goes
- Develops an end-to-end workflow from text extraction and aggregation through geographic features, while also naming advanced imputation and encoding methods.
- Examples
- Demonstrates Tokyo Airbnb price prediction with regex fields, keyword flags, review recency, price aggregation, geographic bins, and proximity distances.
- What you can do after
- Apply it via pandas, regular expressions, aggregation, and scikit-learn KDTree to create and assess tabular features for a prediction task.
Deep Learning with R, Third Edition
Feature engineering is the manual design or transformation of input features into representations suitable for machine learning.
- How it is used
- Makes inputs easier for a learning algorithm to use and helps a small model beat a simple baseline.
- How deep it goes
- Defines hardcoded input transformations, works through a clock example, and gives workflow guidance without a general transformation recipe.
- Examples
- Uses clock-hand coordinates and angles to show how progressively better inputs simplify time prediction.
- What you can do after
- Practice/perform feature selection and design non-learned input transformations using knowledge of the problem.
Read how Deep Learning with R, Third Edition explains feature engineering →
Deep Learning with Python, Third Edition
The manual design or transformation of input variables into representations better suited for a learning algorithm.
- How it is used
- Uses problem knowledge to make model inputs more useful and help a small model beat a baseline.
- How deep it goes
- Defines hardcoded input transformations and works through progressively simpler clock representations without prescribing a universal procedure.
- Examples
- Transforms clock pixels into hand-tip coordinates and then angles, eventually replacing learning with rounding and lookup.
- What you can do after
- Practice/perform it by selecting useful inputs, constructing problem-informed features, and checking whether a model beats a baseline.
Read how Deep Learning with Python, Third Edition explains feature engineering →
Knowledge Graphs and LLMs in Action
Manual feature extraction is the hand-designed selection or computation of features from raw data.
- How it is used
- Converts nodes, relationships, or whole graphs into vectors for downstream classification, link prediction, and other machine-learning tasks.
- How deep it goes
- Descends from manual node metrics to homophily vectors, relationship combinations, metapaths, and DWPC, while contrasting semiautomated and neural alternatives.
- Examples
- Uses karate-club degree features, homophily-based vectors, Node2Vec comparisons, and Hetionet compound–disease metapaths with DWPC.
- What you can do after
- Apply it via NetworkX, tabular classifiers, and graph queries to build interpretable vectors and diagnose representation-driven performance.
Read how Knowledge Graphs and LLMs in Action explains feature engineering →
Natural Language Processing in Action, Second Edition
Feature engineering is the deliberate construction or transformation of input representations to make them useful for predictive modeling.
- How it is used
- Supplies representations for NLP modeling and contrasts hand-designed extraction with features learned by neural networks.
- How deep it goes
- Works through nonlinear transformations, polynomial interaction growth, and a hardcoded convolution, while deferring learned CNN filters.
- Examples
- Uses TF-IDF retrieval, nonlinear input transformations, and a POS-tagged quote processed with a two-point averaging kernel.
- What you can do after
- Build it from scratch with a Python convolution kernel and distinguish specified sequence patterns from patterns a CNN learns.
Read how Natural Language Processing in Action, Second Edition explains feature engineering →
Outlier Detection in Python
Feature engineering is the process of modifying or combining original data attributes to create inputs that are more useful for analysis or modeling.
- How it is used
- Creates detector-ready inputs that expose unusual values, relationships, frequencies, and temporal patterns.
- How deep it goes
- Works through tabular, date, string, and time-series transformations with Python examples, while leaving feature choice dependent on judgment and evaluation.
- Examples
- Uses staff expenses, baseball ratios, phone numbers, date gaps, weekly patterns, and air-quality measurements.
- What you can do after
- Build it from scratch by deriving interpretable columns that help detectors test suspected anomalies in tabular or time-series data.
Read how Outlier Detection in Python explains feature engineering →
Machine Learning System Design
Feature engineering is the process of constructing, transforming, or selecting input features for a machine learning model.
- How it is used
- Turns available data into useful model inputs while guiding baseline and production feature decisions.
- How deep it goes
- Defines an iterative process and gives design checks for computation, availability, reliability, and maintenance, without detailing transformation algorithms.
- Examples
- Uses district-based taxi arrival estimates and model-generated representations to contrast simple and learned inputs.
- What you can do after
- Practice/perform iterative feature design, checking predictive value against serving availability, dependencies, and ongoing reliability.
Read how Machine Learning System Design explains feature engineering →
Effective Data Analysis
Feature engineering is the creation or transformation of variables from existing data to better represent processes relevant to a model.
- How it is used
- Supports exploratory analysis and iterative model refinement by making organizational processes measurable at a useful analytical grain.
- How deep it goes
- Explains aggregation, extraction, encoding, and numerical transformation with examples, while stopping short of detailed algorithmic internals.
- Examples
- Uses rat-sighting counts, weather variables, date components, weekday indicators, dummy coding, boxplots, correlations, and wind-speed transformations.
- What you can do after
- Build it from scratch by deriving, transforming, encoding, and combining variables for explanatory or predictive modeling.
Read how Effective Data Analysis explains feature engineering →
Financial AI in Practice
Feature engineering is the process of constructing informative input variables from raw data for statistical or machine-learning models.
- How it is used
- Turns financial data into signals for credit scoring, fraud detection, market prediction, and news retrieval.
- How deep it goes
- Explains feature choices through concrete financial signals and pipeline requirements, but does not specify a complete implementation in the supplied passages.
- Examples
- Uses fraud velocity and device-link variables, a bond-volatility spread, volume ratios, and news vectors as illustrative artifacts.
- What you can do after
- Practice/perform it by choosing financially meaningful signals and checking whether their timing and underlying relationships support the intended decision.
Read how Financial AI in Practice explains feature engineering →
Delivering Machine Learning Projects: From design to deployment
Feature engineering is the transformation of raw data into informative input variables that represent useful structure for a machine learning model.
- How it is used
- Uses feature engineering as a core ML-pipeline task that prepares raw data before algorithms consume it.
- How deep it goes
- Explains planning, pipeline implementation, bias handling, structural transformations, and augmentation, without deep algorithmic exposition.
- Examples
- Uses calendar sequences, directional values, and temperature plots to illustrate informative representations of cyclical or structured data.
- What you can do after
- Build it from scratch by planning, implementing, reviewing, and augmenting a feature-engineering pipeline.
Sutskever's List
Feature engineering is the process of designing input representations from raw data for use by a learning algorithm.
- How it is used
- Frames handcrafted representations as the input pipeline used by pre-AlexNet vision systems and as a contrast to learned representations.
- How deep it goes
- Explains descriptors, global image representations, classifier inputs, and a failure case, then contrasts them with learned features.
- Examples
- Uses SIFT and HOG examples, a water patch mistaken for a car, and AlexNet’s error-rate comparison with conventional methods.
- What you can do after
- Evaluate how a hand-designed representation can encode useful assumptions yet cause classification errors when its features mislead.
The Art of AI Product Development
Feature engineering is the process of creating, selecting, transforming, and refining input features to improve a machine-learning model's results.
- How it is used
- Improves clustering by turning product behavior and profile data into signals aligned with meaningful customer segments.
- How deep it goes
- Explains selection, enrichment, cleaning, scaling, and iterative assessment, without descending into algorithm implementation.
- Examples
- Uses recommendation reliance, visits, purchases, checkout activity, profile edits, order value, browsing behavior, and product categories.
- What you can do after
- Diagnose/debug weak clusters by removing irrelevant signals, correcting data problems, scaling features, and iterating with domain knowledge.
Read how The Art of AI Product Development explains feature engineering →
Distributed Machine Learning Patterns
Feature engineering is the process of creating, transforming, or selecting input features to improve a machine learning model.
- How it is used
- Improves model inputs by combining overlapping audiovisual features before fitting, while recognizing that deep learning may learn such relationships itself.
- How deep it goes
- Explains derived-feature construction and its performance rationale, then stops short of a procedural feature-engineering workflow.
- Examples
- Uses overlapping Animal and Pet audiovisual features and a 10-class LeNet classification scenario.
- What you can do after
- Evaluate whether combining overlapping audiovisual features is worthwhile before training, especially when exploration threatens single-GPU training time.
Read how Distributed Machine Learning Patterns explains feature engineering →
Digital Twins in Action
Feature engineering is the process of constructing informative input variables from raw data for analysis or machine learning.
- How it is used
- Turns prepared sensor signals into meaningful inputs for physical-system models and predictions.
- How deep it goes
- Works through cyclical, lagged, and domain-specific transformations with pandas examples, while deferring some feature construction to AutoML.
- Examples
- Uses temperature deviations, cyclically encoded hours, three-hour temperature lags, and an indoor–outdoor temperature difference.
- What you can do after
- Build it from scratch with NumPy and pandas transformations, or apply it via AutoML libraries.
Read how Digital Twins in Action explains feature engineering →
Evolutionary Deep Learning: Genetic algorithms and neural networks
Feature engineering is the construction, transformation, or selection of input variables that represent useful information for a machine-learning model.
- How it is used
- Deploys feature construction and selection within automated machine-learning workflows to improve data preparation and model development.
- How deep it goes
- Defines expert-driven feature extraction and its automation, but defers implementation details to later treatment.
- Examples
- Uses expert selection of relevant features and a workflow figure showing feature engineering among AutoML and evolutionary deep-learning steps.
- What you can do after
- Apply feature engineering within AutoML and deep-learning optimization workflows.
Demand Forecasting Best Practices
Feature engineering is the selection, creation, or modification of input features for a machine learning model.
- How it is used
- Improves demand forecasts by supplying models with business-relevant information.
- How deep it goes
- Defines selection, creation, and modification, then gives brainstorming prompts but defers model-parameter tuning.
- Examples
- Uses Black Friday sale status, price changes, average demand, promotions, stockouts, and a typical-inputs table.
- What you can do after
- Apply it by collaboratively brainstorming and collecting business-relevant features for a demand forecast.
Read how Demand Forecasting Best Practices explains feature engineering →
Related concepts
- foundational modeluses · A base or foundation model is a pretrained model that provides learned parameters or capabilities as a starting point for reuse or adaptation to downstream tasks.
- feature selectionhas part · Feature selection is the process of choosing a subset of input features for use in a modeling, analysis, or prediction task.
- data layerpart of · A data layer is an architectural part of a system that organizes access to, management of, or persistence of data for other components.
- pipelinepart of · A pipeline is an ordered or coordinated sequence of connected processing stages or tasks that passes data, work, or results toward an outcome.
- language modeladdressed by · A language model is a computational model that represents patterns in linguistic tokens to estimate sequence probabilities and predict or generate language from context.
- signalproduces · A signal is an observable value, event, or pattern that conveys information about a condition, state, or phenomenon.
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.