Data mining algorithms : explained using R /

• Κύριος συγγραφέας: Cichosz, Pawel (Συγγραφέας)
• Μορφή: Ηλ. βιβλίο
• Γλώσσα: English
• Έκδοση: Chichester, West Sussex ; Malden, MA : John Wiley & Sons Inc., 2015.
• Θέματα: Data mining.; Computer algorithms.; R (Computer program language); MATHEMATICS > Probability & Statistics > General.; Electronic books.

"This book narrows down the scope of data mining by adopting a heavily modeling-oriented perspective"--

Πίνακας περιεχομένων:

• Machine generated contents note: pt. I Preliminaries
• 1. Tasks
• 1.1. Introduction
• 1.1.1. Knowledge
• 1.1.2. Inference
• 1.2. Inductive learning tasks
• 1.2.1. Domain
• 1.2.2. Instances
• 1.2.3. Attributes
• 1.2.4. Target attribute
• 1.2.5. Input attributes
• 1.2.6. Training set
• 1.2.7. Model
• 1.2.8. Performance
• 1.2.9. Generalization
• 1.2.10. Overfitting
• 1.2.11. Algorithms
• 1.2.12. Inductive learning as search
• 1.3. Classification
• 1.3.1. Concept
• 1.3.2. Training set
• 1.3.3. Model
• 1.3.4. Performance
• 1.3.5. Generalization
• 1.3.6. Overfitting
• 1.3.7. Algorithms
• 1.4. Regression
• 1.4.1. Target function
• 1.4.2. Training set
• 1.4.3. Model
• 1.4.4. Performance
• 1.4.5. Generalization
• 1.4.6. Overfitting
• 1.4.7. Algorithms
• 1.5. Clustering
• 1.5.1. Motivation
• 1.5.2. Training set
• 1.5.3. Model
• 1.5.4. Crisp vs. soft clustering
• 1.5.5. Hierarchical clustering
• 1.5.6. Performance
• 1.5.7. Generalization
• 1.5.8. Algorithms.
• 1.5.9. Descriptive vs. predictive clustering
• 1.6. Practical issues
• 1.6.1. Incomplete data
• 1.6.2. Noisy data
• 1.7. Conclusion
• 1.8. Further readings
• References
• 2. Basic statistics
• 2.1. Introduction
• 2.2. Notational conventions
• 2.3. Basic statistics as modeling
• 2.4. Distribution description
• 2.4.1. Continuous attributes
• 2.4.2. Discrete attributes
• 2.4.3. Confidence intervals
• 2.4.4.m-Estimation
• 2.5. Relationship detection
• 2.5.1. Significance tests
• 2.5.2. Continuous attributes
• 2.5.3. Discrete attributes
• 2.5.4. Mixed attributes
• 2.5.5. Relationship detection caveats
• 2.6. Visualization
• 2.6.1. Boxplot
• 2.6.2. Histogram
• 2.6.3. Barplot
• 2.7. Conclusion
• 2.8. Further readings
• References
• pt. II Classification
• 3. Decision trees
• 3.1. Introduction
• 3.2. Decision tree model
• 3.2.1. Nodes and branches
• 3.2.2. Leaves
• 3.2.3. Split types
• 3.3. Growing
• 3.3.1. Algorithm outline.
• 3.3.2. Class distribution calculation
• 3.3.3. Class label assignment
• 3.3.4. Stop criteria
• 3.3.5. Split selection
• 3.3.6. Split application
• 3.3.7.Complete process
• 3.4. Pruning
• 3.4.1. Pruning operators
• 3.4.2. Pruning criterion
• 3.4.3. Pruning control strategy
• 3.4.4. Conversion to rule sets
• 3.5. Prediction
• 3.5.1. Class label prediction
• 3.5.2. Class probability prediction
• 3.6. Weighted instances
• 3.7. Missing value handling
• 3.7.1. Fractional instances
• 3.7.2. Surrogate splits
• 3.8. Conclusion
• 3.9. Further readings
• References
• 4. Naive Bayes classifier
• 4.1. Introduction
• 4.2. Bayes rule
• 4.3. Classification by Bayesian inference
• 4.3.1. Conditional class probability
• 4.3.2. Prior class probability
• 4.3.3. Independence assumption
• 4.3.4. Conditional attribute value probabilities
• 4.3.5. Model construction
• 4.3.6. Prediction
• 4.4. Practical issues
• 4.4.1. Zero and small probabilities.
• 4.4.2. Linear classification
• 4.4.3. Continuous attributes
• 4.4.4. Missing attribute values
• 4.4.5. Reducing naivety
• 4.5. Conclusion
• 4.6. Further readings
• References
• 5. Linear classification
• 5.1. Introduction
• 5.2. Linear representation
• 5.2.1. Inner representation function
• 5.2.2. Outer representation function
• 5.2.3. Threshold representation
• 5.2.4. Logit representation
• 5.3. Parameter estimation
• 5.3.1. Delta rule
• 5.3.2. Gradient descent
• 5.3.3. Distance to decision boundary
• 5.3.4. Least squares
• 5.4. Discrete attributes
• 5.5. Conclusion
• 5.6. Further readings
• References
• 6. Misclassification costs
• 6.1. Introduction
• 6.2. Cost representation
• 6.2.1. Cost matrix
• 6.2.2. Per-class cost vector
• 6.2.3. Instance-specific costs
• 6.3. Incorporating misclassification costs
• 6.3.1. Instance weighting
• 6.3.2. Instance resampling
• 6.3.3. Minimum-cost rule
• 6.3.4. Instance relabeling.
• 6.4. Effects of cost incorporation
• 6.5. Experimental procedure
• 6.6. Conclusion
• 6.7. Further readings
• References
• 7. Classification model evaluation
• 7.1. Introduction
• 7.1.1. Dataset performance
• 7.1.2. Training performance
• 7.1.3. True performance
• 7.2. Performance measures
• 7.2.1. Misclassification error
• 7.2.2. Weighted misclassification error
• 7.2.3. Mean misclassification cost
• 7.2.4. Confusion matrix
• 7.2.5. ROC analysis
• 7.2.6. Probabilistic performance measures
• 7.3. Evaluation procedures
• 7.3.1. Model evaluation vs. modeling procedure evaluation
• 7.3.2. Evaluation caveats
• 7.3.3. Hold-out
• 7.3.4. Cross-validation
• 7.3.5. Leave-one-out
• 7.3.6. Bootstrapping
• 7.3.7. Choosing the right procedure
• 7.3.8. Evaluation procedures for temporal data
• 7.4. Conclusion
• 7.5. Further readings
• References
• pt. III Regression
• 8. Linear regression
• 8.1. Introduction
• 8.2. Linear representation.
• 8.2.1. Parametric representation
• 8.2.2. Linear representation function
• 8.2.3. Nonlinear representation functions
• 8.3. Parameter estimation
• 8.3.1. Mean square error minimization
• 8.3.2. Delta rule
• 8.3.3. Gradient descent
• 8.3.4. Least squares
• 8.4. Discrete attributes
• 8.5. Advantages of linear models
• 8.6. Beyond linearity
• 8.6.1. Generalized linear representation
• 8.6.2. Enhanced representation
• 8.6.3. Polynomial regression
• 8.6.4. Piecewise-linear regression
• 8.7. Conclusion
• 8.8. Further readings
• References
• 9. Regression trees
• 9.1. Introduction
• 9.2. Regression tree model
• 9.2.1. Nodes and branches
• 9.2.2. Leaves
• 9.2.3. Split types
• 9.2.4. Piecewise-constant regression
• 9.3. Growing
• 9.3.1. Algorithm outline
• 9.3.2. Target function summary statistics
• 9.3.3. Target value assignment
• 9.3.4. Stop criteria
• 9.3.5. Split selection
• 9.3.6. Split application
• 9.3.7.Complete process
• 9.4. Pruning.
• 9.4.1. Pruning operators
• 9.4.2. Pruning criterion
• 9.4.3. Pruning control strategy
• 9.5. Prediction
• 9.6. Weighted instances
• 9.7. Missing value handling
• 9.7.1. Fractional instances
• 9.7.2. Surrogate splits
• 9.8. Piecewise linear regression
• 9.8.1. Growing
• 9.8.2. Pruning
• 9.8.3. Prediction
• 9.9. Conclusion
• 9.10. Further readings
• References
• 10. Regression model evaluation
• 10.1. Introduction
• 10.1.1. Dataset performance
• 10.1.2. Training performance
• 10.1.3. True performance
• 10.2. Performance measures
• 10.2.1. Residuals
• 10.2.2. Mean absolute error
• 10.2.3. Mean square error
• 10.2.4. Root mean square error
• 10.2.5. Relative absolute error
• 10.2.6. Coefficient of determination
• 10.2.7. Correlation
• 10.2.8. Weighted performance measures
• 10.2.9. Loss functions
• 10.3. Evaluation procedures
• 10.3.1. Hold-out
• 10.3.2. Cross-validation
• 10.3.3. Leave-one-out
• 10.3.4. Bootstrapping
• 10.3.5. Choosing the right procedure.
• 10.4. Conclusion
• 10.5. Further readings
• References
• pt. IV Clustering
• 11.(Dis)similarity measures
• 11.1. Introduction
• 11.2. Measuring dissimilarity and similarity
• 11.3. Difference-based dissimilarity
• 11.3.1. Euclidean distance
• 11.3.2. Minkowski distance
• 11.3.3. Manhattan distance
• 11.3.4. Canberra distance
• 11.3.5. Chebyshev distance
• 11.3.6. Hamming distance
• 11.3.7. Gower's coefficient
• 11.3.8. Attribute weighting
• 11.3.9. Attribute transformation
• 11.4. Correlation-based similarity
• 11.4.1. Discrete attributes
• 11.4.2. Pearson's correlation similarity
• 11.4.3. Spearman's correlation similarity
• 11.4.4. Cosine similarity
• 11.5. Missing attribute values
• 11.6. Conclusion
• 11.7. Further readings
• References
• 12.k-Centers clustering
• 12.1. Introduction
• 12.1.1. Basic principle
• 12.1.2.(Dis)similarity measures
• 12.2. Algorithm scheme
• 12.2.1. Initialization
• 12.2.2. Stop criteria
• 12.2.3. Cluster formation.
• 12.2.4. Implicit cluster modeling
• 12.2.5. Instantiations
• 12.3.k-Means
• 12.3.1. Center adjustment
• 12.3.2. Minimizing dissimilarity to centers
• 12.4. Beyond means
• 12.4.1.k-Medians
• 12.4.2.k-Medoids
• 12.5. Beyond (fixed) k
• 12.5.1. Multiple runs
• 12.5.2. Adaptive k-centers
• 12.6. Explicit cluster modeling
• 12.7. Conclusion
• 12.8. Further readings
• References
• 13. Hierarchical clustering
• 13.1. Introduction
• 13.1.1. Basic approaches
• 13.1.2.(Dis)similarity measures
• 13.2. Cluster hierarchies
• 13.2.1. Motivation
• 13.2.2. Model representation
• 13.3. Agglomerative clustering
• 13.3.1. Algorithm scheme
• 13.3.2. Cluster linkage
• 13.4. Divisive clustering
• 13.4.1. Algorithm scheme
• 13.4.2. Wrapping a flat clustering algorithm
• 13.4.3. Stop criteria
• 13.5. Hierarchical clustering visualization
• 13.6. Hierarchical clustering prediction
• 13.6.1. Cutting cluster hierarchies
• 13.6.2. Cluster membership assignment.
• 13.7. Conclusion
• 13.8. Further readings
• References
• 14. Clustering model evaluation
• 14.1. Introduction
• 14.1.1. Dataset performance
• 14.1.2. Training performance
• 14.1.3. True performance
• 14.2. Per-cluster quality measures
• 14.2.1. Diameter
• 14.2.2. Separation
• 14.2.3. Isolation
• 14.2.4. Silhouette width
• 14.2.5. Davies
• Bouldin Index
• 14.3. Overall quality measures
• 14.3.1. Dunn Index
• 14.3.2. Average Davies
• Bouldin Index
• 14.3.3.C Index
• 14.3.4. Average silhouette width
• 14.3.5. Loglikelihood
• 14.4. External quality measures
• 14.4.1. Misclassification error
• 14.4.2. Rand Index
• 14.4.3. General relationship detection measures
• 14.5. Using quality measures
• 14.6. Conclusion
• 14.7. Further readings
• References
• pt. V Getting Better Models
• 15. Model ensembles
• 15.1. Introduction
• 15.2. Model committees
• 15.3. Base models
• 15.3.1. Different training sets
• 15.3.2. Different algorithms.
• 15.3.3. Different parameter setups
• 15.3.4. Algorithm randomization
• 15.3.5. Base model diversity
• 15.4. Model aggregation
• 15.4.1. Voting/Averaging
• 15.4.2. Probability averaging
• 15.4.3. Weighted voting/averaging
• 15.4.4. Using as attributes
• 15.5. Specific ensemble modeling algorithms
• 15.5.1. Bagging
• 15.5.2. Stacking
• 15.5.3. Boosting
• 15.5.4. Random forest
• 15.5.5. Random Naive Bayes
• 15.6. Quality of ensemble predictions
• 15.7. Conclusion
• 15.8. Further readings
• References
• 16. Kernel methods
• 16.1. Introduction
• 16.2. Support vector machines
• 16.2.1. Classification margin
• 16.2.2. Maximum-margin hyperplane
• 16.2.3. Primal form
• 16.2.4. Dual form
• 16.2.5. Soft margin
• 16.3. Support vector regression
• 16.3.1. Regression tube
• 16.3.2. Primal form
• 16.3.3. Dual form
• 16.4. Kernel trick
• 16.5. Kernel functions
• 16.5.1. Linear kernel
• 16.5.2. Polynomial kernel
• 16.5.3. Radial kernel
• 16.5.4. Sigmoid kernel.
• 16.6. Kernel prediction
• 16.7. Kernel-based algorithms
• 16.7.1. Kernel-based SVM
• 16.7.2. Kernel-based SVR
• 16.8. Conclusion
• 16.9. Further readings
• References
• 17. Attribute transformation
• 17.1. Introduction
• 17.2. Attribute transformation task
• 17.2.1. Target task
• 17.2.2. Target attribute
• 17.2.3. Transformed attribute
• 17.2.4. Training set
• 17.2.5. Modeling transformations
• 17.2.6. Nonmodeling transformations
• 17.3. Simple transformations
• 17.3.1. Standardization
• 17.3.2. Normalization
• 17.3.3. Aggregation
• 17.3.4. Imputation
• 17.3.5. Binary encoding
• 17.4. Multiclass encoding
• 17.4.1. Encoding and decoding functions
• 17.4.2.1-ok-k encoding
• 17.4.3. Error-correcting encoding
• 17.4.4. Effects of multiclass encoding
• 17.5. Conclusion
• 17.6. Further readings
• References
• 18. Discretization
• 18.1. Introduction
• 18.2. Discretization task
• 18.2.1. Motivation
• 18.2.2. Task definition.
• 18.2.3. Discretization as modeling
• 18.2.4. Discretization quality
• 18.3. Unsupervised discretization
• 18.3.1. Equal-width intervals
• 18.3.2. Equal-frequency intervals
• 18.3.3. Nonmodeling discretization
• 18.4. Supervised discretization
• 18.4.1. Pure-class discretization
• 18.4.2. Bottom-up discretization
• 18.4.3. Top-down discretization
• 18.5. Effects of discretization
• 18.6. Conclusion
• 18.7. Further readings
• References
• 19. Attribute selection
• 19.1. Introduction
• 19.2. Attribute selection task
• 19.2.1. Motivation
• 19.2.2. Task definition
• 19.2.3. Algorithms
• 19.3. Attribute subset search
• 19.3.1. Search task
• 19.3.2. Initial state
• 19.3.3. Search operators
• 19.3.4. State selection
• 19.3.5. Stop criteria
• 19.4. Attribute selection filters
• 19.4.1. Simple statistical filters
• 19.4.2. Correlation-based filters
• 19.4.3. Consistency-based filters
• 19.4.4. RELIEF
• 19.4.5. Random forest
• 19.4.6. Cutoff criteria.
• 19.4.7. Filter-driven search
• 19.5. Attribute selection wrappers
• 19.5.1. Subset evaluation
• 19.5.2. Wrapper attribute selection
• 19.6. Effects of attribute selection
• 19.7. Conclusion
• 19.8. Further readings
• References
• 20. Case studies
• 20.1. Introduction
• 20.1.1. Datasets
• 20.1.2. Packages
• 20.1.3. Auxiliary functions
• 20.2. Census income
• 20.2.1. Data loading and preprocessing
• 20.2.2. Default model
• 20.2.3. Incorporating misclassification costs
• 20.2.4. Pruning
• 20.2.5. Attribute selection
• 20.2.6. Final models
• 20.3.Communities and crime
• 20.3.1. Data loading
• 20.3.2. Data quality
• 20.3.3. Regression trees
• 20.3.4. Linear models
• 20.3.5. Attribute selection
• 20.3.6. Piecewise-linear models
• 20.4. Cover type
• 20.4.1. Data loading and preprocessing
• 20.4.2. Class imbalance
• 20.4.3. Decision trees
• 20.4.4. Class rebalancing
• 20.4.5. Multiclass encoding
• 20.4.6. Final classification models
• 20.4.7. Clustering.
• 20.5. Conclusion
• 20.6. Further readings
• References
• Closing
• A. Notation
• A.1. Attribute values
• A.2. Data subsets
• A.3. Probabilities
• B.R packages
• B.1. CRAN packages
• B.2. DMR packages
• B.3. Installing packages
• References
• C. Datasets.