Preprocessing
Preprocessing is the preparation or transformation of data, content, or other inputs into a form suitable for a subsequent computational operation.
Deep Learning with R, Third Edition
Data preprocessing is the preparation of raw data by reshaping, scaling, encoding, or otherwise converting it into a model-compatible form.
- How it is used
- Prepares raw inputs and targets as model-compatible tensors before training and evaluation.
- How deep it goes
- Works through reshaping, scaling, tokenization, dataset pipelines, and detection targets; leaves domain-specific choices to practical examples.
- Examples
- Uses MNIST pixels, JPEG image directories, IMDb reviews, house-price features, and grid-aligned bounding boxes.
- What you can do after
- Apply it via Keras and tf.data to reshape or encode data, normalize values, and feed prepared batches to a model.
Read how Deep Learning with R, Third Edition explains preprocessing →
Data Without Labels
Data preprocessing, or data preparation, organizes and transforms raw data into a form suitable for analysis or model training.
- How it is used
- Positions preprocessing between data discovery and exploratory analysis or modeling, making raw data usable for the intended task.
- How deep it goes
- Details common data problems and task-specific treatments, but the supplied passages do not spell out the code for each operation.
- Examples
- Uses retail invoices requiring conversion for association rules and text cleaned through stop-word removal, normalization, and tokenization.
- What you can do after
- Practice/perform data-quality inspection and choose preparation steps suited to a dataset and its modeling task.
Software Testing with Generative AI
Data preprocessing is the preparation and transformation of data into a form suitable for model training or fine-tuning.
- How it is used
- Prepares business-domain documentation and code for instruction-based fine-tuning.
- How deep it goes
- Explains schemas, parsing choices, prompt encoding, and tokenization, but stops short of detailing tokenizer internals.
- Examples
- Uses JSONL instruction-output pairs and Java controller and test-code fragments to contrast coherent records with isolated lines.
- What you can do after
- Apply it via Axolotl to tokenize a prepared dataset before launching a fine-tuning run.
Read how Software Testing with Generative AI explains preprocessing →
Optimization Algorithms: AI techniques for design, planning, and control problems
A preprocessing phase is an initial computation that transforms or augments data before subsequent queries or operations.
- How it is used
- Speeds shortest-path queries by preparing an augmented graph before bidirectional search.
- How deep it goes
- Works through node ordering, contraction, shortcut decisions, and the later query phase, with a Python listing for predetermined order.
- Examples
- Uses a road-network query comparison, contraction diagrams, and a NetworkX listing with predetermined node order.
- What you can do after
- Apply it via NetworkX to prepare a contraction hierarchy with predetermined node order and inspect its effect on shortest-path queries.
Deep Learning with Python, Third Edition
The transformation of raw data into a form suitable for input to a model.
- How it is used
- Makes raw inputs and targets suitable for neural-network training.
- How deep it goes
- Works through vectorization, normalization, text transformations, and detection targets; leaves many domain-specific techniques to their practical examples.
- Examples
- Uses IMDb reviews, MNIST pixels, house-price features, and grid-based bounding-box targets.
- What you can do after
- Practice/perform converting text and numeric features into model-ready inputs and preparing grid-aligned detection targets.
Read how Deep Learning with Python, Third Edition explains preprocessing →
Grokking Machine Learning
Data preprocessing is the transformation of raw data into a form suitable for computational analysis or model training.
- How it is used
- Prepares raw emails, labels, images, and tabular records for model training and evaluation.
- How deep it goes
- Demonstrates concrete transformations and an end-to-end workflow, while stopping short of exhaustive data-cleaning rules.
- Examples
- Uses spam emails, circular two-class data, MNIST images, and the Titanic dataset as preprocessing examples.
- What you can do after
- Apply preprocessing with Pandas and Keras by transforming text, labels, images, and tabular data into model-ready representations.
Ensemble Methods for Machine Learning
Preprocessing is the transformation of raw data into a representation suitable for machine-learning algorithms.
- How it is used
- Converts text, categorical values, and numeric measurements into representations suitable for machine-learning algorithms.
- How deep it goes
- Walks through cleaning, tokenization, lemmatization, vectorization, vocabulary filtering, TF-IDF, encoding, and standardization with implementation examples.
- Examples
- Uses movie reviews, a term-document matrix, TF-IDF extraction, categorical variables, and the AutoMPG regression data set.
- What you can do after
- Apply it via NLTK and scikit-learn to clean, encode, vectorize, reweight, and standardize data before training.
Read how Ensemble Methods for Machine Learning explains preprocessing →
Distributed Machine Learning Patterns
The transformation and preparation of raw data into a form suitable for downstream computation or model training.
- How it is used
- Prepares ingested images for model training and converts serving requests into model-compatible inputs.
- How deep it goes
- Demonstrates scaling, casting, caching, shuffling, decoding, resizing, and serving signatures without detailing transformation internals.
- Examples
- Uses MNIST normalization, a Fashion-MNIST tf.data pipeline, and JPEG-byte decoding for image inference.
- What you can do after
- Apply it via TensorFlow and tf.data to normalize training data and define preprocessing in a serving signature.
Read how Distributed Machine Learning Patterns explains preprocessing →
Grokking AI Algorithms, Second Edition
Data preparation is the process of cleaning, formatting, and transforming raw data into a form suitable for analysis or model development.
- How it is used
- Prepares collected data for model training, with choices shaped by the question and the data type.
- How deep it goes
- Works through concrete tabular cleanup decisions and stepwise text and image cleaning, without specifying a universal preprocessing recipe.
- Examples
- Uses a diamond dataset with missing values and categorical encoding, plus noisy text and a scraped pet-blog image.
- What you can do after
- Practice/perform data preparation by choosing cleanup and encoding steps for tabular data and identifying noise to remove from text or images.
Read how Grokking AI Algorithms, Second Edition explains preprocessing →
Machine Learning Platform Engineering
Data preparation is the process of cleaning, transforming, organizing, and otherwise making data suitable for model development.
- How it is used
- Anchors model development by converting data into a reproducible input for training, evaluation, and later production workflows.
- How deep it goes
- Breaks preparation into downloading, splitting, notebook exploration, and Kubeflow component construction, while deferring broader pipeline stages.
- Examples
- Uses ID-card detection and movie recommendation projects, with dataset download and training, test, and validation splits.
- What you can do after
- Build it from scratch as composable Kubeflow components assembled into an executable, reproducible preparation pipeline.
Read how Machine Learning Platform Engineering explains preprocessing →
Machine Learning System Design
Preprocessing is the transformation and preparation of data before it is supplied to a machine learning model for training or inference.
- How it is used
- Prepares collected data for aggregation, feature calculation, model training, or prediction.
- How deep it goes
- Explains filtering, aggregation, repair, joining, and modality-specific transformations, but does not detail individual algorithms.
- Examples
- Uses retail transactions, stock, promotions, weather, competitor prices, images, and text as preprocessing examples.
- What you can do after
- Build it from scratch as a staged transformation flow that produces cleaner, joined, feature-ready inputs.
Read how Machine Learning System Design explains preprocessing →
Generative AI in Action
Data preprocessing is the process of cleaning, transforming, and organizing raw data into a form suitable for analysis or model processing.
- How it is used
- Prepares enterprise data for RAG, helping improve generated answers and inform document chunking.
- How deep it goes
- Lists cleaning and transformation operations and places them in a data pipeline, without an executable preprocessing implementation.
- Examples
- Contrasts a simple capital-city question with a multifaceted economic question to motivate chunk-size evaluation.
- What you can do after
- Apply it via Unstructured to create customized preprocessing pipelines, or use LangChain and SK for chunking with overlap.
Starting Data Analytics with Generative AI and Python
Data preprocessing is the transformation and preparation of raw data into a form suitable for analysis or modeling.
- How it is used
- Prepares business data for descriptive analysis, visualization, statistical inference, and modeling.
- How deep it goes
- Explains quality checks, table merging, missing-value handling, type conversion, and derived variables, then stops before detailed modeling.
- Examples
- Uses merged order datasets, delivery-time differences, missing-value decisions, converted timestamps, summary statistics, and visualizations.
- What you can do after
- Apply preprocessing by validating quality, combining relevant tables, cleaning fields, and creating variables for analysis.
Read how Starting Data Analytics with Generative AI and Python explains preprocessing →
Machine Learning for Tabular Data: XGBoost, Deep Learning, and AI
Data preprocessing is the transformation and preparation of raw data before it is supplied to a machine learning model.
- How it is used
- Prepares stored data and features for machine-learning model training.
- How deep it goes
- Provides a worked inspection and cleanup workflow, stopping short of specifying final transformations for every problematic column.
- Examples
- Uses a Kuala Lumpur real estate dataset, DataFrame statistics, unique-value counts, dimensions, and missing-value reports.
- What you can do after
- Apply DataFrame inspection to classify features, quantify missing data, and prioritize column cleanup before training.
Read how Machine Learning for Tabular Data: XGBoost, Deep Learning, and AI explains preprocessing →
Graph Neural Networks in Action
Data preprocessing is the transformation and preparation of data into a form suitable for model training or inference.
- How it is used
- Prepares graph-structured data for GNN training and inference through representation, splitting, sampling, batching, and loading.
- How deep it goes
- Defines graph conversion, splitting, sampling, and batching, while deferring deeper preprocessing detail to appendix B.
- Examples
- Demonstrates NetworkX-to-PyG conversion, tensor assignment, indexed splits, masks, and full-graph or batched model input.
- What you can do after
- Apply preprocessing via PyTorch Geometric and NetworkX to produce graph data suitable for GNN training.
Read how Graph Neural Networks in Action explains preprocessing →
Building Reliable AI Systems
The cleaning and transformation of raw data before it is supplied to a machine-learning process.
- How it is used
- Prepares usable examples for LLM fine-tuning before model selection and training.
- How deep it goes
- Specifies artifact removal, notation preservation, and consistent example formatting, but does not provide an automated cleaning pipeline.
- Examples
- Shows technical-support Customer/Agent JSONL and legal User/AI prompt-completion examples.
- What you can do after
- Practice/perform preprocessing by removing confusing artifacts while preserving domain notation and consistent role-labeled examples.
Read how Building Reliable AI Systems explains preprocessing →
Outlier Detection in Python
The preparation and transformation of raw data into a form suitable for analysis or machine-learning models.
- How it is used
- Prepares heterogeneous tabular data for outlier detection and supports iterative detector development and evaluation.
- How deep it goes
- Explains cleaning, encoding, binning, scaling, imputation, feature work, and ensemble variation, but does not derive their algorithms.
- Examples
- Uses NaN counts, median-filled Strikeouts values, and IQR-based removal of extreme numeric records.
- What you can do after
- Apply it via tabular-data operations to produce detector-compatible features and construct preprocessing variants for ensembles.
Read how Outlier Detection in Python explains preprocessing →
The Art of AI Product Development
Preprocessing is the transformation and preparation of raw data before it is supplied to a machine-learning model.
- How it is used
- Prepares raw data for mathematical computation by producing model-ready numerical representations.
- How deep it goes
- Defines numerical transformation, cleaning, and information loss, but stops short of detailing embedding implementation.
- Examples
- Uses one-hot word vectors, word embeddings, and numerical financial or statistical data as examples.
- What you can do after
- Evaluate preprocessing choices by checking model compatibility, modality expertise, and task-relevant information loss.
Read how The Art of AI Product Development explains preprocessing →
Deep Learning with PyTorch, Second Edition
Data preparation performed before model execution, here separating image and prompt encoding from the main model run.
- How it is used
- Prepares images, masks, prompts, and captions so pretrained or trainable models can consume them.
- How deep it goes
- Demonstrates concrete transformations and embedding preparation but defers detailed model-training and encoding internals.
- Examples
- Uses BLIP captions, SAM point prompts, CT images, masks, SegformerImageProcessor, and TensorBoard loss tracking.
- What you can do after
- Apply preprocessing via SegformerImageProcessor or SAM’s image-and-prompt workflow before model inference or training.
Read how Deep Learning with PyTorch, Second Edition explains preprocessing →
Knowledge Graphs and LLMs in Action
Data preparation is the process of selecting, cleaning, transforming, and organizing data for computational use.
- How it is used
- Prepares financial, recommendation, and knowledge-graph data for graph learning or structured analysis.
- How deep it goes
- Explains multi-stage identifier, tensor, feature, label, schema, and ontology transformations with implementation examples.
- Examples
- Uses financial transactions, illicit-node labels, movie genres, user ratings, medical documents, and ontology mappings.
- What you can do after
- Builds numerical graph inputs and structured knowledge-graph data through mappings, encoding, ingestion, and relationship generation.
Read how Knowledge Graphs and LLMs in Action explains preprocessing →
Effective Conversational AI
Data preparation is the process of transforming collected data into a suitable form for testing or model training.
- How it is used
- Prepares production-log utterances for testing and documents for retrieval-augmented generation searches.
- How deep it goes
- Outlines filtering, scrubbing, annotation, chunking, embedding, and storage without detailing each step’s implementation.
- Examples
- Uses logged utterances, PDFs, tables, semantic chunks, intent labels, and ideal response annotations as examples.
- What you can do after
- Apply it to prepare test sets from logs or searchable document data for RAG.
Read how Effective Conversational AI explains preprocessing →
Evolutionary Deep Learning: Genetic algorithms and neural networks
The process of cleaning, transforming, and organizing data for machine-learning analysis or training.
- How it is used
- Positions data preparation at the start of the AI/ML workflow before feature engineering and model development.
- How deep it goes
- Frames preparation as costly and automatable but defers concrete cleaning procedures to later treatment.
- Examples
- Uses the AutoML workflow diagram and names Azure Machine Learning, AutoKeras, and AutoTorch as automation platforms.
- What you can do after
- Evaluate where data preparation fits in an automated machine-learning workflow and why it matters for model performance.
Read how Evolutionary Deep Learning: Genetic algorithms and neural networks explains preprocessing →
Related concepts
- 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.
- data cleaninghas part · Data cleaning is the process of detecting and correcting, removing, transforming, or otherwise handling inaccurate, incomplete, inconsistent, duplicate, or unsuitable data.
- standardhas part · A standard is an established rule, specification, or convention that defines a consistent reference for practice, representation, behavior, or evaluation.
- indexinghas part · Indexing is the organization or use of positions, keys, or lookup structures to identify and retrieve selected data or elements.
- machine learning workflowpart of · A machine learning workflow is an organized sequence of activities for formulating, developing, evaluating, deploying, operating, and maintaining a machine learning system.
- categorical encodinghas part · Categorical encoding transforms discrete category values into numerical representations, such as separate binary indicators or vectors with one active position per category.
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.