cross-validation
Rotate training subsets
Cross-validation is a statistical technique for testing how well a predictive model will generalize to new data.
The data is split into subsets; the model trains on some subsets and is tested on others, rotating through until each subset has served as the test set.
This guards against overfitting—a model that performs well on training data but poorly on fresh data.
We used ten-fold cross-validation to evaluate the classifier.
Arranged in stacked layers
- holdoutOne refusing to participate
- nestedContained within another
- repeatedDone or said again
- bootstrappingSelf-starting without external input
jackknifing · resampling · partitioning · folds · iterations · splits · sampling · Monte Carlo
Excess adaptation to samples
- underfittingModel too simple
- generalizationDeriving principles from examples
- regularizationPenalty preventing model complexity
- learning curveRate of skill acquisition
Degree of correctness
- precisionAccuracy of measurement
- recallMemory retrieval act
- metricsStandards for assessment
- error rateMeasure of mistakes made
AUC · performance · benchmarking · scoring
Data used to teach models
- foldsBends upon itself
- partitionsBarriers dividing a space
- splitsDivides into parts
- sample sizeQuantity of observations taken
stratified
Arranged in stacked layers
bootstrapping · jackknifing · resampling · partitioning · folds · iterations · splits · sampling · Monte Carlo · blocked
overfitting
Excess adaptation to samples
- underfittingModel too simple
- generalizationDeriving principles from examples
- regularizationPenalty preventing model complexity
accuracy
Degree of correctness
error rate · AUC · performance · benchmarking · scoring
training set
Data used to teach models
- foldsBends upon itself
- partitionsBarriers dividing a space
- splitsDivides into parts
- data sciencealgorithm, data analytics, data, machine learning, dataset
- research methodsexperiment, hypothesis, deduction, experiments, induction
- forecastingforecast, prediction, prognosis, outlook, estimate
- validationapproval, accuracy, precision, specificity, selectivity
- machine learningneural networks, deep learning, backpropagation, underfitting, CNNs
- Statistical MethodsANOVA, confidence intervals, correlation, covariance, hypothesis testing
cross, valid, validate, invalidate, validation, validity, validated, invalidating, validates