book

Machine Learning with R - Fourth Edition

by Brett Lantz

May 2023

Intermediate to advanced

762 pages

21h

English

Packt Publishing

Read now

Unlock full access

Who this book is forWhat this book coversWhat you need for this bookGet in touch
The origins of machine learningUses and abuses of machine learningMachine learning successesThe limits of machine learningMachine learning ethicsHow machines learnData storageAbstractionGeneralizationEvaluationMachine learning in practiceTypes of input dataTypes of machine learning algorithmsMatching input data to algorithmsMachine learning with RInstalling R packagesLoading and unloading R packagesInstalling RStudioWhy R and why R now?Summary
R data structuresVectorsFactorsListsData framesMatrices and arraysManaging data with RSaving, loading, and removing R data structuresImporting and saving datasets from CSV filesImporting common dataset formats using RStudioExploring and understanding dataExploring the structure of dataExploring numeric featuresMeasuring the central tendency – mean and medianMeasuring spread – quartiles and the five-number summaryVisualizing numeric features – boxplotsVisualizing numeric features – histogramsUnderstanding numeric data – uniform and normal distributionsMeasuring spread – variance and standard deviationExploring categorical featuresMeasuring the central tendency – the modeExploring relationships between featuresVisualizing relationships – scatterplotsExamining relationships – two-way cross-tabulationsSummary
Understanding nearest neighbor classificationThe k-NN algorithmMeasuring similarity with distanceChoosing an appropriate kPreparing data for use with k-NNWhy is the k-NN algorithm lazy?Example – diagnosing breast cancer with the k-NN algorithmStep 1 – collecting dataStep 2 – exploring and preparing the dataTransformation – normalizing numeric dataData preparation – creating training and test datasetsStep 3 – training a model on the dataStep 4 – evaluating model performanceStep 5 – improving model performanceTransformation – z-score standardizationTesting alternative values of kSummary
Understanding Naive BayesBasic concepts of Bayesian methodsUnderstanding probabilityUnderstanding joint probabilityComputing conditional probability with Bayes’ theoremThe Naive Bayes algorithmClassification with Naive BayesThe Laplace estimatorUsing numeric features with Naive BayesExample – filtering mobile phone spam with the Naive Bayes algorithmStep 1 – collecting dataStep 2 – exploring and preparing the dataData preparation – cleaning and standardizing text dataData preparation – splitting text documents into wordsData preparation – creating training and test datasetsVisualizing text data – word cloudsData preparation – creating indicator features for frequent wordsStep 3 – training a model on the dataStep 4 – evaluating model performanceStep 5 – improving model performanceSummary
Understanding decision treesDivide and conquerThe C5.0 decision tree algorithmChoosing the best splitPruning the decision treeExample – identifying risky bank loans using C5.0 decision treesStep 1 – collecting dataStep 2 – exploring and preparing the dataData preparation – creating random training and test datasetsStep 3 – training a model on the dataStep 4 – evaluating model performanceStep 5 – improving model performanceBoosting the accuracy of decision treesMaking some mistakes cost more than othersUnderstanding classification rulesSeparate and conquerThe 1R algorithmThe RIPPER algorithmRules from decision treesWhat makes trees and rules greedy?Example – identifying poisonous mushrooms with rule learnersStep 1 – collecting dataStep 2 – exploring and preparing the dataStep 3 – training a model on the dataStep 4 – evaluating model performanceStep 5 – improving model performanceSummary
Understanding regressionSimple linear regressionOrdinary least squares estimationCorrelationsMultiple linear regressionGeneralized linear models and logistic regressionExample – predicting auto insurance claims costs using linear regressionStep 1 – collecting dataStep 2 – exploring and preparing the dataExploring relationships between features – the correlation matrixVisualizing relationships between features – the scatterplot matrixStep 3 – training a model on the dataStep 4 – evaluating model performanceStep 5 – improving model performanceModel specification – adding nonlinear relationshipsModel specification – adding interaction effectsPutting it all together – an improved regression modelMaking predictions with a regression modelGoing further – predicting insurance policyholder churn with logistic regressionUnderstanding regression trees and model treesAdding regression to treesExample – estimating the quality of wines with regression trees and model treesStep 1 – collecting dataStep 2 – exploring and preparing the dataStep 3 – training a model on the dataVisualizing decision treesStep 4 – evaluating model performanceMeasuring performance with the mean absolute errorStep 5 – improving model performanceSummary
Understanding neural networksFrom biological to artificial neuronsActivation functionsNetwork topologyThe number of layersThe direction of information travelThe number of nodes in each layerTraining neural networks with backpropagationExample – modeling the strength of concrete with ANNsStep 1 – collecting dataStep 2 – exploring and preparing the dataStep 3 – training a model on the dataStep 4 – evaluating model performanceStep 5 – improving model performanceUnderstanding support vector machinesClassification with hyperplanesThe case of linearly separable dataThe case of nonlinearly separable dataUsing kernels for nonlinear spacesExample – performing OCR with SVMsStep 1 – collecting dataStep 2 – exploring and preparing the dataStep 3 – training a model on the dataStep 4 – evaluating model performanceStep 5 – improving model performanceChanging the SVM kernel functionIdentifying the best SVM cost parameterSummary
Understanding association rulesThe Apriori algorithm for association rule learningMeasuring rule interest – support and confidenceBuilding a set of rules with the Apriori principleExample – identifying frequently purchased groceries with association rulesStep 1 – collecting dataStep 2 – exploring and preparing the dataData preparation – creating a sparse matrix for transaction dataVisualizing item support – item frequency plotsVisualizing the transaction data – plotting the sparse matrixStep 3 – training a model on the dataStep 4 – evaluating model performanceStep 5 – improving model performanceSorting the set of association rulesTaking subsets of association rulesSaving association rules to a file or data frameUsing the Eclat algorithm for greater efficiencySummary
Understanding clusteringClustering as a machine learning taskClusters of clustering algorithmsThe k-means clustering algorithmUsing distance to assign and update clusters Choosing the appropriate number of clustersFinding teen market segments using k-means clusteringStep 1 – collecting dataStep 2 – exploring and preparing the dataData preparation – dummy coding missing valuesData preparation – imputing the missing valuesStep 3 – training a model on the dataStep 4 – evaluating model performanceStep 5 – improving model performanceSummary

Measuring performance for classificationUnderstanding a classifier’s predictionsA closer look at confusion matricesUsing confusion matrices to measure performanceBeyond accuracy – other measures of performanceThe kappa statisticThe Matthews correlation coefficientSensitivity and specificityPrecision and recallThe F-measureVisualizing performance tradeoffs with ROC curvesComparing ROC curvesThe area under the ROC curveCreating ROC curves and computing AUC in REstimating future performanceThe holdout methodCross-validationBootstrap samplingSummary
What makes a successful machine learning practitioner?What makes a successful machine learning model?Avoiding obvious predictionsConducting fair evaluationsConsidering real-world impactsBuilding trust in the modelPutting the “science” in data scienceUsing R Notebooks and R MarkdownPerforming advanced data explorationConstructing a data exploration roadmapEncountering outliers: a real-world pitfallExample – using ggplot2 for visual data explorationSummary
Performing feature engineeringThe role of human and machineThe impact of big data and deep learningFeature engineering in practiceHint 1: Brainstorm new featuresHint 2: Find insights hidden in textHint 3: Transform numeric rangesHint 4: Observe neighbors’ behaviorHint 5: Utilize related rowsHint 6: Decompose time seriesHint 7: Append external dataExploring R’s tidyverseMaking tidy table structures with tibblesReading rectangular files faster with readr and readxlPreparing and piping data with dplyrTransforming text with stringrCleaning dates with lubridateSummary
The challenge of high-dimension dataApplying feature selectionFilter methodsWrapper methods and embedded methodsExample – Using stepwise regression for feature selectionExample – Using Boruta for feature selectionPerforming feature extractionUnderstanding principal component analysisExample – Using PCA to reduce highly dimensional social media dataMaking use of sparse dataIdentifying sparse dataExample – Remapping sparse categorical dataExample – Binning sparse numeric dataHandling missing dataUnderstanding types of missing dataPerforming missing value imputationSimple imputation with missing value indicatorsMissing value patternsThe problem of imbalanced dataSimple strategies for rebalancing dataGenerating a synthetic balanced dataset with SMOTEExample – Applying the SMOTE algorithm in RConsidering whether balanced is always betterSummary
Tuning stock models for better performanceDetermining the scope of hyperparameter tuningExample – using caret for automated tuningCreating a simple tuned modelCustomizing the tuning processImproving model performance with ensemblesUnderstanding ensemble learningPopular ensemble-based algorithmsBaggingBoostingRandom forestsGradient boostingExtreme gradient boosting with XGBoostWhy are tree-based ensembles so popular?Stacking models for meta-learningUnderstanding model stacking and blendingPractical methods for blending and stacking in RSummary
Practical applications of deep learningBeginning with deep learningChoosing appropriate tasks for deep learningThe TensorFlow and Keras deep learning frameworksUnderstanding convolutional neural networksTransfer learning and fine tuningExample – classifying images using a pre-trained CNN in RUnsupervised learning and big dataRepresenting highly dimensional concepts as embeddingsUnderstanding word embeddingsExample – using word2vec for understanding text in RVisualizing highly dimensional dataThe limitations of using PCA for big data visualizationUnderstanding the t-SNE algorithmExample – visualizing data’s natural clusters with t-SNEAdapting R to handle large datasetsQuerying data in SQL databasesThe tidy approach to managing database connectionsUsing a database backend for dplyr with dbplyrDoing work faster with parallel processingMeasuring R’s execution timeEnabling parallel processing in RTaking advantage of parallel with foreach and doParallelTraining and evaluating models in parallel with caretUtilizing specialized hardware and algorithmsParallel computing with MapReduce concepts via Apache SparkLearning via distributed and scalable algorithms with H2OGPU computingSummary

Content preview from Machine Learning with R - Fourth Edition

10 Evaluating Model Performance

When only the wealthy could afford education, tests and exams were not used to evaluate students. Instead, tests evaluated the teachers for parents who wanted to know whether their children learned enough to justify the instructors’ wages. Obviously, this is different today. Now, such evaluations are used to distinguish between high-achieving and low-achieving students, filtering them into careers and other opportunities.

Given the significance of this process, a great deal of effort is invested in developing accurate student assessments. Fair assessments have a large number of questions that cover a wide breadth of topics and reward true knowledge over lucky guesses. A good assessment also requires students to ...