book

Mastering Predictive Analytics with R

by Rui Miguel Forte

June 2015

Beginner to intermediate

414 pages

11h 26m

English

Packt Publishing

Read now

Unlock full access

Support files, eBooks, discount offers, and moreWhy subscribe?Free access for Packt account holders
What this book covers

Downloading the example codeErrataPiracyQuestions
ModelsLearning from dataThe core components of a modelOur first model: k-nearest neighbors
Supervised, unsupervised, semi-supervised, and reinforcement learning modelsParametric and nonparametric modelsRegression and classification modelsReal-time and batch machine learning models
Defining the model's objectiveCollecting the dataPicking a modelPreprocessing the dataExploratory data analysisFeature transformationsEncoding categorical featuresMissing dataOutliersRemoving problematic featuresFeature engineering and dimensionality reductionTraining and assessing the modelRepeating with different models and final model selectionDeploying the model
Assessing regression modelsAssessing classification modelsAssessing binary classification models
Introduction to linear regressionAssumptions of linear regression
Estimating the regression coefficients
Predicting CPU performancePredicting the price of used cars
Residual analysisSignificance tests for linear regressionPerformance metrics for linear regressionComparing different regression modelsTest set performance
MulticollinearityOutliers
Ridge regressionLeast absolute shrinkage and selection operator (lasso)Implementing regularization in R
Classifying with linear regression
Generalized linear modelsInterpreting coefficients in logistic regressionAssumptions of logistic regressionMaximum likelihood estimation
Model devianceTest set performance
Multinomial logistic regressionPredicting glass typeOrdinal logistic regressionPredicting wine quality
The biological neuron
Gradient descent and local minimaThe perceptron algorithmLinear separationThe logistic neuron
Training multilayer perceptron networks
Evaluating multilayer perceptrons for regression
Receiver operating characteristic curves
Maximal margin classification
Inner products
The intuition for tree models
Classification and regression treesCART regression treesTree pruningMissing dataRegression model treesCART classification treesC5.0
Tuning model parameters in CART treesVariable importance in tree modelsRegression model trees in action
BaggingMargins and out-of-bag observationsPredicting complex skill learning with baggingPredicting heart disease with baggingLimitations of bagging
AdaBoost
Limitations of boosting
The importance of variables in random forests
A little graph theory
Predicting the sentiment of movie reviews
Fundamental concepts of time seriesTime series summary functions
White noiseFitting a white noise time seriesRandom walkFitting a random walk
Moving average modelsAutoregressive modelsAutoregressive moving average models
Autoregressive integrated moving average modelsAutoregressive conditional heteroscedasticity modelsGeneralized autoregressive heteroscedasticity models
An overview of topic modeling
The Dirichlet distributionThe generative processFitting an LDA model
Model stabilityFinding the number of topicsTopic distributionsWord distributionsLDA extensions
Rating matrixMeasuring user similarity
User-based collaborative filteringItem-based collaborative filtering
Evaluating binary top-N recommendationsEvaluating non-binary top-N recommendationsEvaluating individual predictions

Content preview from Mastering Predictive Analytics with R

Types of models

With a broad idea of the basic components of a model, we are ready to explore some of the common distinctions that modelers use to categorize different models.

Supervised, unsupervised, semi-supervised, and reinforcement learning models

We've already looked at the iris data set, which consisted of four features and one output variable, namely the species variable. Having the output variable available for all the observations in the training data is the defining characteristic of the supervised learning setting, which represents the most frequent scenario encountered. In a nutshell, the advantage of training a model under the supervised learning setting is that we have the correct answer that we should be predicting for the data points ...