Feature Engineering and Selection : A Practical Approach for Predictive Models (Hardcover)

1. IntroductionA Simple ExampleImportant ConceptsA More Complex ExampleFeature SelectionAn Outline of the BookComputing 2. Illustrative Example: Predicting Risk of Ischemic StrokeSplittingPreprocessingExplorationPredictive Modeling Across SetsOther ConsiderationsComputing 3. A Review of the Predictive Modeling ProcessIllustrative Example: OkCupid Profile DataMeasuring PerformanceData SplittingResamplingTuning Parameters and OverfittingModel Optimization and TuningComparing Models Using the Training SetFeature Engineering Without OverfittingSummaryComputing 4. Exploratory VisualizationsIntroduction to the Chicago Train Ridership DataVisualizations for Numeric Data: Exploring Train Ridership DataVisualizations for Categorical Data: Exploring the OkCupid DataPost Modeling Exploratory VisualizationsSummaryComputing 5. Encoding Categorical PredictorsCreating Dummy Variables for Unordered CategoriesEncoding Predictors with Many CategoriesApproaches for Novel CategoriesSupervised Encoding MethodsEncodings for Ordered DataCreating Features from Text DataFactors versus Dummy Variables in Tree-Based ModelsSummaryComputing 6. Engineering Numeric PredictorsTransformationsMany TransformationsMany: Many TransformationsSummaryComputing 7. Detecting Interaction EffectsGuiding Principles in the Search for InteractionsPractical ConsiderationsThe Brute-Force Approach to Identifying Predictive InteractionsApproaches when Complete Enumeration is Practically ImpossibleOther Potentially Useful ToolsSummaryComputing 8. Handling Missing DataUnderstanding the Nature and Severity of Missing InformationModels that are Resistant to Missing ValuesDeletion of DataEncoding MissingnessImputation methodsSpecial CasesSummaryComputing 9. Working with Profile DataIllustrative Data: Pharmaceutical Manufacturing MonitoringWhat are the Experimental Unit and the Unit of Prediction?Reducing BackgroundReducing Other NoiseExploiting CorrelationImpacts of Data Processing on ModelingSummaryComputing 10. Feature Selection OverviewGoals of Feature SelectionClasses of Feature Selection MethodologiesEffect of Irrelevant FeaturesOverfitting to Predictors and External ValidationA Case StudyNext StepsComputing 11. Greedy Search MethodsIllustrative Data: Predicting Parkinson’s DiseaseSimple FiltersRecursive Feature EliminationStepwise SelectionSummaryComputing 12. Global Search MethodsNaive Bayes ModelsSimulated AnnealingGenetic AlgorithmsTest Set ResultsSummaryComputing