TASC Data Science and Digital Solutions flashcards

32 cards. Read the term and the definition here. The deck at the top of this page is optional.

CRISP-DM

A six-phase process model for data projects: business understanding, data understanding, data preparation, modeling, evaluation, and deployment, with iterative revisiting of phases.

Supervised learning

A machine learning approach that trains a model on labeled input-output pairs so it can predict the target for new, unseen inputs.

Unsupervised learning

A machine learning approach that finds patterns, groups, or structure in unlabeled data, such as clustering or dimensionality reduction.

Overfitting

A model fits training data noise so closely that it generalizes poorly; symptoms include low training error and high test error.

Underfitting

A model is too simple to capture the underlying pattern, producing high error on both training and test data.

Bias-variance tradeoff

The tension between a model's systematic error (bias) and its sensitivity to training data variation (variance); complexity reduces bias and raises variance.

Precision

The fraction of predicted positives that are truly positive: TP divided by TP plus FP. It measures positive prediction correctness.

Recall

The fraction of actual positives correctly identified: TP divided by TP plus FN. It measures the ability to find all positives.

F1 score

The harmonic mean of precision and recall, computed as 2 times precision times recall divided by precision plus recall.

Confusion matrix

A table of predicted versus actual class counts with true positives, true negatives, false positives, and false negatives, the basis for classification metrics.

Cross-validation

A resampling technique that splits data into k folds, training on k-1 and testing on the remaining fold, then averaging results.

Data leakage

Occurrence when information from outside the training set, such as test data or future values, contaminates model training and inflates performance estimates.

Feature engineering

The process of creating, transforming, or selecting input variables to improve model performance, such as scaling, encoding, or deriving ratios.

Regularization

A technique that penalizes model complexity, for example L1 or L2 penalties, to reduce overfitting and improve generalization.

Clustering

An unsupervised method that groups similar observations, such as k-means or hierarchical clustering, without predefined labels.

Train-test split

Partitioning a dataset into a training portion for fitting a model and a held-out portion for unbiased evaluation of generalization.

Correlation versus causation

A statistical association between variables does not establish that one causes the other; confounding and bias can produce spurious relationships.

ROC-AUC

The area under the receiver operating characteristic curve; it summarizes a classifier's ability to rank positives above negatives across all thresholds.

Data pipeline

An automated sequence of steps that ingests, validates, transforms, stores, and serves data for analytics or machine learning.

Data quality dimensions

Attributes used to assess data fitness, including completeness, accuracy, consistency, timeliness, validity, and uniqueness.

GDPR

The European Union's General Data Protection Regulation governing personal data processing through principles such as lawfulness, minimization, purpose limitation, and accountability.

Data minimization

A privacy principle requiring that only data adequate, relevant, and necessary for a stated purpose be collected and processed.

Pseudonymization

Replacing direct identifiers with artificial keys so data cannot be attributed without extra information; the data remains personal data under GDPR.

Anonymization

Irreversibly removing or altering identifying information so that individuals can no longer be re-identified, taking the data outside GDPR scope.

RBAC

Role-based access control assigns permissions to roles rather than individuals, and users receive permissions through assigned roles.

ABAC

Attribute-based access control grants access by evaluating attributes of the user, resource, action, and environment against policy rules.

Symmetric encryption

Encryption using a single shared secret key for both encryption and decryption, such as AES; it is fast but requires secure key distribution.

Asymmetric encryption

Encryption using a public-private key pair, such as RSA or ECC, enabling secure key exchange and digital signatures without sharing a secret.

Cloud service models

IaaS provides virtualized infrastructure, PaaS provides managed platforms and runtimes, and SaaS provides complete applications delivered over the network.

Shared responsibility model

A cloud security division in which the provider secures the cloud infrastructure while the customer secures data, identities, and configurations.

REST API

An architectural style for web services using stateless HTTP requests and resource-oriented URIs, typically with GET, POST, PUT, and DELETE methods.

Model drift

Degradation of model performance over time as real-world data distributions change, requiring monitoring, retraining, or recalibration.