Skip to main content

Machine Learning (Apuntes)

Machine Learning

El machine learning ha experimentado un crecimiento significativo durante la 煤ltima d茅cada. Este desarrollo se atribuye a tres factores fundamentales: el incremento sustancial en la disponibilidad de datos (Big Data), la evoluci贸n de las capacidades computacionales y el perfeccionamiento de algoritmos avanzados. En la actualidad, el machine learning constituye un elemento transformador en diversos sectores: desde aplicaciones m茅dicas para el diagn贸stico de enfermedades, hasta la optimizaci贸n de estrategias financieras. Su capacidad anal铆tica y de procesamiento de datos lo posiciona como un recurso esencial para la planificaci贸n estrat茅gica, la optimizaci贸n de procesos y el desarrollo de soluciones personalizadas.

驴Cu谩ndo Implementar Machine Learning?

La implementaci贸n del machine learning resulta particularmente efectiva en los siguientes contextos:

  • Disponibilidad de datos a gran escala: La eficacia del modelo se incrementa proporcionalmente con la cantidad de datos relevantes disponibles.
  • Presencia de relaciones complejas entre variables: En situaciones donde la multiplicidad de variables dificulta la definici贸n de reglas convencionales.
  • Necesidad de adaptaci贸n din谩mica: Los sistemas de machine learning permiten una optimizaci贸n continua mediante la incorporaci贸n de nueva informaci贸n.
  • Requerimientos de automatizaci贸n avanzada: Facilita la ejecuci贸n de tareas complejas, desde el an谩lisis visual hasta la generaci贸n de pron贸sticos predictivos.

馃搶 Cuadro

Modelos y cuando usarlos
Tipo Problema_tipico Ventajas Cuando_usarlo
Regressi贸n Valores num茅ricos Simplicidad Relaciones lineales
脕rboles / Decision Tree Clasificaci贸n, regresi贸n Interpretabilidad Datos tabulares
Ensambles Clasificaci贸n, regresi贸n Precisi贸n Alto rendimiento, Kaggle
Deep Learning Im谩genes, texto, audio Modelos complejos Datos grandes y no estructurados
Reducci贸n de Dim. Visualizaci贸n, preprocesamiento Mejora eficiencia Datos con muchas variables
Bayesianos Clasificaci贸n r谩pida Velocidad Texto, spam detection
Regularizaci贸n Evitar overfitting Generalizaci贸n Modelos lineales con muchas variables
Instance-Based Clasificaci贸n Simple, no requiere entrenamiento Pocos datos y relaciones claras
Clustering Agrupamiento no supervisado Descubrir estructuras ocultas Segmentaci贸n sin etiquetas
Rule-Based Interpretabilidad L贸gica clara Reglas conocidas, decisiones explicables
Fuente: Elaboraci贸n propia

Modelos de Machine Learning

Modelos disponibles en la paqueter铆a caret en R.

Enlace: https://topepo.github.io/caret/model-training-and-tuning.html

Modelos de Machine Learning
Disponibles en la paqueter铆a caret
Model Method Value Type Libraries Tunning Parameters
AdaBoost Classification Trees adaboost Classification fastAdaboost nIter, method
AdaBoost.M1 AdaBoost.M1 Classification adabag, plyr mfinal, maxdepth, coeflearn
Adaptive-Network-Based Fuzzy Inference System ANFIS Regression frbs num.labels, max.iter
Adaptive Mixture Discriminant Analysis amdai Classification adaptDA model
Adjacent Categories Probability Model for Ordinal Data vglmAdjCat Classification VGAM parallel, link
Bagged AdaBoost AdaBag Classification adabag, plyr mfinal, maxdepth
Bagged CART treebag Classification, Regression ipred, plyr, e1071 None
Bagged FDA using gCV Pruning bagFDAGCV Classification earth degree
Bagged Flexible Discriminant Analysis bagFDA Classification earth, mda degree, nprune
Bagged Logic Regression logicBag Classification, Regression logicFS nleaves, ntrees
Bagged MARS bagEarth Classification, Regression earth nprune, degree
Bagged MARS using gCV Pruning bagEarthGCV Classification, Regression earth degree
Bagged Model bag Classification, Regression caret vars
Bayesian Additive Regression Trees bartMachine Classification, Regression bartMachine num_trees, k, alpha, beta, nu
Bayesian Generalized Linear Model bayesglm Classification, Regression arm None
Bayesian Regularized Neural Networks brnn Regression brnn neurons
Bayesian Ridge Regression bridge Regression monomvn None
Bayesian Ridge Regression (Model Averaged) blassoAveraged Regression monomvn None
Binary Discriminant Analysis binda Classification binda lambda.freqs
Boosted Classification Trees ada Classification ada, plyr iter, maxdepth, nu
Boosted Generalized Additive Model gamboost Classification, Regression mboost, plyr, import mstop, prune
Boosted Generalized Linear Model glmboost Classification, Regression plyr, mboost mstop, prune
Boosted Linear Model BstLm Classification, Regression bst, plyr mstop, nu
Boosted Logistic Regression LogitBoost Classification caTools nIter
Boosted Smoothing Spline bstSm Classification, Regression bst, plyr mstop, nu
Boosted Tree blackboost Classification, Regression party, mboost, plyr, partykit mstop, maxdepth
Boosted Tree bstTree Classification, Regression bst, plyr mstop, maxdepth, nu
C4.5-like Trees J48 Classification RWeka C, M
C5.0 C5.0 Classification C50, plyr trials, model, winnow
CART rpart Classification, Regression rpart cp
CART rpart1SE Classification, Regression rpart None
CART rpart2 Classification, Regression rpart maxdepth
CART or Ordinal Responses rpartScore Classification rpartScore, plyr cp, split, prune
CHi-squared Automated Interaction Detection chaid Classification CHAID alpha2, alpha3, alpha4
Conditional Inference Random Forest cforest Classification, Regression party mtry
Conditional Inference Tree ctree Classification, Regression party mincriterion
Conditional Inference Tree ctree2 Classification, Regression party maxdepth, mincriterion
Continuation Ratio Model for Ordinal Data vglmContRatio Classification VGAM parallel, link
Cost-Sensitive C5.0 C5.0Cost Classification C50, plyr trials, model, winnow, cost
Cost-Sensitive CART rpartCost Classification rpart, plyr cp, Cost
Cubist cubist Regression Cubist committees, neighbors
Cumulative Probability Model for Ordinal Data vglmCumulative Classification VGAM parallel, link
DeepBoost deepboost Classification deepboost num_iter, tree_depth, beta, lambda, loss_type
Diagonal Discriminant Analysis dda Classification sparsediscrim model, shrinkage
Distance Weighted Discrimination with Polynomial Kernel dwdPoly Classification kerndwd lambda, qval, degree, scale
Distance Weighted Discrimination with Radial Basis Function Kernel dwdRadial Classification kernlab, kerndwd lambda, qval, sigma
Dynamic Evolving Neural-Fuzzy Inference System DENFIS Regression frbs Dthr, max.iter
Elasticnet enet Regression elasticnet fraction, lambda
Ensembles of Generalized Linear Models randomGLM Classification, Regression randomGLM maxInteractionOrder
eXtreme Gradient Boosting xgbDART Classification, Regression xgboost, plyr nrounds, max_depth, eta, gamma, subsample, colsample_bytree, rate_drop, skip_drop, min_child_weight
eXtreme Gradient Boosting xgbLinear Classification, Regression xgboost nrounds, lambda, alpha, eta
eXtreme Gradient Boosting xgbTree Classification, Regression xgboost, plyr nrounds, max_depth, eta, gamma, colsample_bytree, min_child_weight, subsample
Extreme Learning Machine elm Classification, Regression elmNN nhid, actfun
Factor-Based Linear Discriminant Analysis RFlda Classification HiDimDA q
Flexible Discriminant Analysis fda Classification earth, mda degree, nprune
Fuzzy Inference Rules by Descent Method FIR.DM Regression frbs num.labels, max.iter
Fuzzy Rules Using Chi's Method FRBCS.CHI Classification frbs num.labels, type.mf
Fuzzy Rules Using Genetic Cooperative-Competitive Learning and Pittsburgh FH.GBML Classification frbs max.num.rule, popu.size, max.gen
Fuzzy Rules Using the Structural Learning Algorithm on Vague Environment SLAVE Classification frbs num.labels, max.iter, max.gen
Fuzzy Rules via MOGUL GFS.FR.MOGUL Regression frbs max.gen, max.iter, max.tune
Fuzzy Rules via Thrift GFS.THRIFT Regression frbs popu.size, num.labels, max.gen
Fuzzy Rules with Weight Factor FRBCS.W Classification frbs num.labels, type.mf
Gaussian Process gaussprLinear Classification, Regression kernlab None
Gaussian Process with Polynomial Kernel gaussprPoly Classification, Regression kernlab degree, scale
Gaussian Process with Radial Basis Function Kernel gaussprRadial Classification, Regression kernlab sigma
Generalized Additive Model using LOESS gamLoess Classification, Regression gam span, degree
Generalized Additive Model using Splines bam Classification, Regression mgcv select, method
Generalized Additive Model using Splines gam Classification, Regression mgcv select, method
Generalized Additive Model using Splines gamSpline Classification, Regression gam df
Generalized Linear Model glm Classification, Regression None
Generalized Linear Model with Stepwise Feature Selection glmStepAIC Classification, Regression MASS None
Generalized Partial Least Squares gpls Classification gpls K.prov
Genetic Lateral Tuning and Rule Selection of Linguistic Fuzzy Systems GFS.LT.RS Regression frbs popu.size, num.labels, max.gen
glmnet glmnet Classification, Regression glmnet, Matrix alpha, lambda
glmnet glmnet_h2o Classification, Regression h2o alpha, lambda
Gradient Boosting Machines gbm_h2o Classification, Regression h2o ntrees, max_depth, min_rows, learn_rate, col_sample_rate
Greedy Prototype Selection protoclass Classification proxy, protoclass eps, Minkowski
Heteroscedastic Discriminant Analysis hda Classification hda gamma, lambda, newdim
High-Dimensional Regularized Discriminant Analysis hdrda Classification sparsediscrim gamma, lambda, shrinkage_type
High Dimensional Discriminant Analysis hdda Classification HDclassif threshold, model
Hybrid Neural Fuzzy Inference System HYFIS Regression frbs num.labels, max.iter
Independent Component Regression icr Regression fastICA n.comp
k-Nearest Neighbors kknn Classification, Regression kknn kmax, distance, kernel
k-Nearest Neighbors knn Classification, Regression k
L2 Regularized Linear Support Vector Machines with Class Weights svmLinearWeights2 Classification LiblineaR cost, Loss, weight
L2 Regularized Support Vector Machine (dual) with Linear Kernel svmLinear3 Classification, Regression LiblineaR cost, Loss
Learning Vector Quantization lvq Classification class size, k
Least Angle Regression lars Regression lars fraction
Least Angle Regression lars2 Regression lars step
Least Squares Support Vector Machine lssvmLinear Classification kernlab tau
Least Squares Support Vector Machine with Polynomial Kernel lssvmPoly Classification kernlab degree, scale, tau
Least Squares Support Vector Machine with Radial Basis Function Kernel lssvmRadial Classification kernlab sigma, tau
Linear Discriminant Analysis lda Classification MASS None
Linear Discriminant Analysis lda2 Classification MASS dimen
Linear Discriminant Analysis with Stepwise Feature Selection stepLDA Classification klaR, MASS maxvar, direction
Linear Distance Weighted Discrimination dwdLinear Classification kerndwd lambda, qval
Linear Regression lm Regression intercept
Linear Regression with Backwards Selection leapBackward Regression leaps nvmax
Linear Regression with Forward Selection leapForward Regression leaps nvmax
Linear Regression with Stepwise Selection leapSeq Regression leaps nvmax
Linear Regression with Stepwise Selection lmStepAIC Regression MASS None
Linear Support Vector Machines with Class Weights svmLinearWeights Classification e1071 cost, weight
Localized Linear Discriminant Analysis loclda Classification klaR k
Logic Regression logreg Classification, Regression LogicReg treesize, ntrees
Logistic Model Trees LMT Classification RWeka iter
Maximum Uncertainty Linear Discriminant Analysis Mlda Classification HiDimDA None
Mixture Discriminant Analysis mda Classification mda subclasses
Model Averaged Naive Bayes Classifier manb Classification bnclassify smooth, prior
Model Averaged Neural Network avNNet Classification, Regression nnet size, decay, bag
Model Rules M5Rules Regression RWeka pruned, smoothed
Model Tree M5 Regression RWeka pruned, smoothed, rules
Monotone Multi-Layer Perceptron Neural Network monmlp Classification, Regression monmlp hidden1, n.ensemble
Multi-Layer Perceptron mlp Classification, Regression RSNNS size
Multi-Layer Perceptron mlpWeightDecay Classification, Regression RSNNS size, decay
Multi-Layer Perceptron, multiple layers mlpWeightDecayML Classification, Regression RSNNS layer1, layer2, layer3, decay
Multi-Layer Perceptron, with multiple layers mlpML Classification, Regression RSNNS layer1, layer2, layer3
Multi-Step Adaptive MCP-Net msaenet Classification, Regression msaenet alphas, nsteps, scale
Multilayer Perceptron Network by Stochastic Gradient Descent mlpSGD Classification, Regression FCNN4R, plyr size, l2reg, lambda, learn_rate, momentum, gamma, minibatchsz, repeats
Multilayer Perceptron Network with Dropout mlpKerasDropout Classification, Regression keras size, dropout, batch_size, lr, rho, decay, activation
Multilayer Perceptron Network with Dropout mlpKerasDropoutCost Classification keras size, dropout, batch_size, lr, rho, decay, cost, activation
Multilayer Perceptron Network with Weight Decay mlpKerasDecay Classification, Regression keras size, lambda, batch_size, lr, rho, decay, activation
Multilayer Perceptron Network with Weight Decay mlpKerasDecayCost Classification keras size, lambda, batch_size, lr, rho, decay, cost, activation
Multivariate Adaptive Regression Spline earth Classification, Regression earth nprune, degree
Multivariate Adaptive Regression Splines gcvEarth Classification, Regression earth degree
Naive Bayes naive_bayes Classification naivebayes laplace, usekernel, adjust
Naive Bayes nb Classification klaR fL, usekernel, adjust
Naive Bayes Classifier nbDiscrete Classification bnclassify smooth
Naive Bayes Classifier with Attribute Weighting awnb Classification bnclassify smooth
Nearest Shrunken Centroids pam Classification pamr threshold
Negative Binomial Generalized Linear Model glm.nb Regression MASS link
Neural Network mxnet Classification, Regression mxnet layer1, layer2, layer3, learning.rate, momentum, dropout, activation
Neural Network mxnetAdam Classification, Regression mxnet layer1, layer2, layer3, dropout, beta1, beta2, learningrate, activation
Neural Network neuralnet Regression neuralnet layer1, layer2, layer3
Neural Network nnet Classification, Regression nnet size, decay
Neural Networks with Feature Extraction pcaNNet Classification, Regression nnet size, decay
Non-Convex Penalized Quantile Regression rqnc Regression rqPen lambda, penalty
Non-Informative Model null Classification, Regression None
Non-Negative Least Squares nnls Regression nnls None
Oblique Random Forest ORFlog Classification obliqueRF mtry
Oblique Random Forest ORFpls Classification obliqueRF mtry
Oblique Random Forest ORFridge Classification obliqueRF mtry
Oblique Random Forest ORFsvm Classification obliqueRF mtry
Optimal Weighted Nearest Neighbor Classifier ownn Classification snn K
Ordered Logistic or Probit Regression polr Classification MASS method
Parallel Random Forest parRF Classification, Regression e1071, randomForest, foreach, import mtry
partDSA partDSA Classification, Regression partDSA cut.off.growth, MPD
Partial Least Squares kernelpls Classification, Regression pls ncomp
Partial Least Squares pls Classification, Regression pls ncomp
Partial Least Squares simpls Classification, Regression pls ncomp
Partial Least Squares widekernelpls Classification, Regression pls ncomp
Partial Least Squares Generalized Linear Models plsRglm Classification, Regression plsRglm nt, alpha.pvals.expli
Patient Rule Induction Method PRIM Classification supervisedPRIM peel.alpha, paste.alpha, mass.min
Penalized Discriminant Analysis pda Classification mda lambda
Penalized Discriminant Analysis pda2 Classification mda df
Penalized Linear Discriminant Analysis PenalizedLDA Classification penalizedLDA, plyr lambda, K
Penalized Linear Regression penalized Regression penalized lambda1, lambda2
Penalized Logistic Regression plr Classification stepPlr lambda, cp
Penalized Multinomial Regression multinom Classification nnet decay
Penalized Ordinal Regression ordinalNet Classification ordinalNet, plyr alpha, criteria, link, lambda, modeltype, family
Polynomial Kernel Regularized Least Squares krlsPoly Regression KRLS lambda, degree
Prediction Rule Ensembles pre Classification, Regression pre sampfrac, maxdepth, learnrate, mtry, use.grad, penalty.par.val
Principal Component Analysis pcr Regression pls ncomp
Projection Pursuit Regression ppr Regression nterms
Quadratic Discriminant Analysis qda Classification MASS None
Quadratic Discriminant Analysis with Stepwise Feature Selection stepQDA Classification klaR, MASS maxvar, direction
Quantile Random Forest qrf Regression quantregForest mtry
Quantile Regression Neural Network qrnn Regression qrnn n.hidden, penalty, bag
Quantile Regression with LASSO penalty rqlasso Regression rqPen lambda
Radial Basis Function Kernel Regularized Least Squares krlsRadial Regression KRLS, kernlab lambda, sigma
Radial Basis Function Network rbf Classification, Regression RSNNS size
Radial Basis Function Network rbfDDA Classification, Regression RSNNS negativeThreshold
Random Ferns rFerns Classification rFerns depth
Random Forest ordinalRF Classification e1071, ranger, dplyr, ordinalForest nsets, ntreeperdiv, ntreefinal
Random Forest ranger Classification, Regression e1071, ranger, dplyr mtry, splitrule, min.node.size
Random Forest Rborist Classification, Regression Rborist predFixed, minNode
Random Forest rf Classification, Regression randomForest mtry
Random Forest by Randomization extraTrees Classification, Regression extraTrees mtry, numRandomCuts
Random Forest Rule-Based Model rfRules Classification, Regression randomForest, inTrees, plyr mtry, maxdepth
Regularized Discriminant Analysis rda Classification klaR gamma, lambda
Regularized Linear Discriminant Analysis rlda Classification sparsediscrim estimator
Regularized Logistic Regression regLogistic Classification LiblineaR cost, loss, epsilon
Regularized Random Forest RRF Classification, Regression randomForest, RRF mtry, coefReg, coefImp
Regularized Random Forest RRFglobal Classification, Regression RRF mtry, coefReg
Relaxed Lasso relaxo Regression relaxo, plyr lambda, phi
Relevance Vector Machines with Linear Kernel rvmLinear Regression kernlab None
Relevance Vector Machines with Polynomial Kernel rvmPoly Regression kernlab scale, degree
Relevance Vector Machines with Radial Basis Function Kernel rvmRadial Regression kernlab sigma
Ridge Regression ridge Regression elasticnet lambda
Ridge Regression with Variable Selection foba Regression foba k, lambda
Robust Linear Discriminant Analysis Linda Classification rrcov None
Robust Linear Model rlm Regression MASS intercept, psi
Robust Mixture Discriminant Analysis rmda Classification robustDA K, model
Robust Quadratic Discriminant Analysis QdaCov Classification rrcov None
Robust Regularized Linear Discriminant Analysis rrlda Classification rrlda lambda, hp, penalty
Robust SIMCA RSimca Classification rrcovHD None
ROC-Based Classifier rocc Classification rocc xgenes
Rotation Forest rotationForest Classification rotationForest K, L
Rotation Forest rotationForestCp Classification rpart, plyr, rotationForest K, L, cp
Rule-Based Classifier JRip Classification RWeka NumOpt, NumFolds, MinWeights
Rule-Based Classifier PART Classification RWeka threshold, pruned
Self-Organizing Maps xyf Classification, Regression kohonen xdim, ydim, user.weights, topo
Semi-Naive Structure Learner Wrapper nbSearch Classification bnclassify k, epsilon, smooth, final_smooth, direction
Shrinkage Discriminant Analysis sda Classification sda diagonal, lambda
SIMCA CSimca Classification rrcov, rrcovHD None
Simplified TSK Fuzzy Rules FS.HGD Regression frbs num.labels, max.iter
Single C5.0 Ruleset C5.0Rules Classification C50 None
Single C5.0 Tree C5.0Tree Classification C50 None
Single Rule Classification OneR Classification RWeka None
Sparse Distance Weighted Discrimination sdwd Classification sdwd lambda, lambda2
Sparse Linear Discriminant Analysis sparseLDA Classification sparseLDA NumVars, lambda
Sparse Mixture Discriminant Analysis smda Classification sparseLDA NumVars, lambda, R
Sparse Partial Least Squares spls Classification, Regression spls K, eta, kappa
Spike and Slab Regression spikeslab Regression spikeslab, plyr vars
Stabilized Linear Discriminant Analysis slda Classification ipred None
Stabilized Nearest Neighbor Classifier snn Classification snn lambda
Stacked AutoEncoder Deep Neural Network dnn Classification, Regression deepnet layer1, layer2, layer3, hidden_dropout, visible_dropout
Stochastic Gradient Boosting gbm Classification, Regression gbm, plyr n.trees, interaction.depth, shrinkage, n.minobsinnode
Subtractive Clustering and Fuzzy c-Means Rules SBC Regression frbs r.a, eps.high, eps.low
Supervised Principal Component Analysis superpc Regression superpc threshold, n.components
Support Vector Machines with Boundrange String Kernel svmBoundrangeString Classification, Regression kernlab length, C
Support Vector Machines with Class Weights svmRadialWeights Classification kernlab sigma, C, Weight
Support Vector Machines with Exponential String Kernel svmExpoString Classification, Regression kernlab lambda, C
Support Vector Machines with Linear Kernel svmLinear Classification, Regression kernlab C
Support Vector Machines with Linear Kernel svmLinear2 Classification, Regression e1071 cost
Support Vector Machines with Polynomial Kernel svmPoly Classification, Regression kernlab degree, scale, C
Support Vector Machines with Radial Basis Function Kernel svmRadial Classification, Regression kernlab sigma, C
Support Vector Machines with Radial Basis Function Kernel svmRadialCost Classification, Regression kernlab C
Support Vector Machines with Radial Basis Function Kernel svmRadialSigma Classification, Regression kernlab sigma, C
Support Vector Machines with Spectrum String Kernel svmSpectrumString Classification, Regression kernlab length, C
The Bayesian lasso blasso Regression monomvn sparsity
The lasso lasso Regression elasticnet fraction
Tree-Based Ensembles nodeHarvest Classification, Regression nodeHarvest maxinter, mode
Tree Augmented Naive Bayes Classifier tan Classification bnclassify score, smooth
Tree Augmented Naive Bayes Classifier Structure Learner Wrapper tanSearch Classification bnclassify k, epsilon, smooth, final_smooth, sp
Tree Augmented Naive Bayes Classifier with Attribute Weighting awtan Classification bnclassify score, smooth
Tree Models from Genetic Algorithms evtree Classification, Regression evtree alpha
Variational Bayesian Multinomial Probit Regression vbmpRadial Classification vbmp estimateTheta
Wang and Mendel Fuzzy Rules WM Regression frbs num.labels, type.mf
Weighted Subspace Random Forest wsrf Classification wsrf mtry
Fuente: Elaboraci贸n propia