Index
A
- A/B testing, Testing Production Systems
- accuracy, Evaluating the Model, Relation to accuracy
- acknowledgments, From Andreas
- adjusted rand index (ARI), Evaluating clustering with ground truth
- agglomerative clustering
- algorithm chains and pipelines, Algorithm Chains and Pipelines-Summary and Outlook
- algorithm parameter, Estimating complexity in neural networks
- algorithms (see also models; problem solving)
- evaluating, Generalization, Overfitting, and Underfitting
- minimal code to apply to algorithm, Summary and Outlook
- sample datasets, Some Sample Datasets-Some Sample Datasets
- scaling
- MinMaxScaler, Preprocessing data for SVMs, Applying Data Transformations-The Effect of Preprocessing on Supervised Learning, DBSCAN, Interactions and Polynomials, Building Pipelines, Grid-Searching Which Model To Use
- Normalizer, Different Kinds of Preprocessing
- RobustScaler, Different Kinds of Preprocessing
- StandardScaler, Tuning neural networks, Different Kinds of Preprocessing, Scaling Training and Test Data the Same Way, Applying PCA to the cancer dataset for visualization, Eigenfaces for feature extraction, DBSCAN-Evaluating clustering without ground truth, Convenient Pipeline Creation with make_pipeline-Grid-Searching Which Model To Use
- supervised, classification
- decision trees, Decision Trees-Strengths, weaknesses, and parameters
- gradient boosting, Gradient boosted regression trees (gradient boosting machines)-Gradient boosted regression trees (gradient boosting machines), Uncertainty Estimates from Classifiers, Uncertainty in Multiclass Classification
- k-nearest neighbors, k-Nearest Neighbors-Strengths, weaknesses, and parameters
- kernelized support vector machines, Kernelized Support Vector Machines-Strengths, weaknesses, and parameters
- linear SVMs, Linear models for classification
- logistic regression, Linear models for classification
- naive Bayes, Naive Bayes Classifiers-Strengths, weaknesses, and parameters
- neural networks, Neural Networks (Deep Learning)-Estimating complexity in neural networks
- random forests, Building random forests-Strengths, weaknesses, and parameters
- supervised, regression
- decision trees, Decision Trees-Strengths, weaknesses, and parameters
- gradient boosting, Gradient boosted regression trees (gradient boosting machines)-Gradient boosted regression trees (gradient boosting machines)
- k-nearest neighbors, k-neighbors regression
- Lasso, Lasso-Lasso
- linear regression (OLS), Linear regression (aka ordinary least squares), Binning, Discretization, Linear Models, and Trees-Interactions and Polynomials
- neural networks, Neural Networks (Deep Learning)-Estimating complexity in neural networks
- random forests, Building random forests-Strengths, weaknesses, and parameters
- Ridge, Ridge regression-Lasso, Strengths, weaknesses, and parameters, Tuning neural networks, Interactions and Polynomials, Univariate Nonlinear Transformations, Using Pipelines in Grid Searches, Grid-Searching Preprocessing Steps and Model Parameters-Grid-Searching Preprocessing Steps and Model Parameters
- unsupervised, clustering
- unsupervised, manifold learning
- unsupervised, signal decomposition
- alpha parameter in linear models, Ridge regression
- Anaconda, Installing scikit-learn
- analysis of variance (ANOVA), Univariate Statistics
- area under the curve (AUC), Receiver operating characteristics (ROC) and AUC-Receiver operating characteristics (ROC) and AUC
- attributions, Using Code Examples
- average precision, Precision-recall curves and ROC curves
B
- bag-of-words representation
- BernoulliNB, Naive Bayes Classifiers
- bigrams, Bag-of-Words with More Than One Word (n-Grams)
- binary classification, Classification and Regression, Linear models for classification, Metrics for Binary Classification-Receiver operating characteristics (ROC) and AUC
- binning, Applying PCA to the cancer dataset for visualization, Binning, Discretization, Linear Models, and Trees-Binning, Discretization, Linear Models, and Trees
- bootstrap samples, Building random forests
- Boston Housing dataset, Some Sample Datasets
- boundary points, DBSCAN
- Bunch objects, Some Sample Datasets
- business metric, Keep the End Goal in Mind, Approaching a Machine Learning Problem
C
- C parameter in SVC, Tuning SVM parameters
- calibration, Taking uncertainty into account
- cancer dataset, Some Sample Datasets
- categorical features
- categorical variables (see categorical features)
- chaining (see algorithm chains and pipelines)
- class labels, Classification and Regression
- classification problems
- classifiers
- DecisionTreeClassifier, Controlling complexity of decision trees, Imbalanced datasets
- DecisionTreeRegressor, Controlling complexity of decision trees, Feature importance in trees
- KNeighborsClassifier, Building Your First Model: k-Nearest Neighbors-Summary and Outlook, k-Neighbors classification-k-neighbors regression
- KNeighborsRegressor, k-neighbors regression-Linear models for regression
- LinearSVC, Linear models for classification-Linear models for classification, Linear models for multiclass classification, Strengths, weaknesses, and parameters, Naive Bayes Classifiers
- LogisticRegression, Linear models for classification-Linear models for classification, Strengths, weaknesses, and parameters, Summary and Outlook, Cross-Validation in scikit-learn, Imbalanced datasets, Accessing Attributes in a Pipeline inside GridSearchCV, Bag-of-Words for Movie Reviews-Advanced Tokenization, Stemming, and Lemmatization
- MLPClassifier, The neural network model-Estimating complexity in neural networks
- naive Bayes, Naive Bayes Classifiers-Strengths, weaknesses, and parameters
- SVC, Linear models for classification, Tuning SVM parameters, Applying Data Transformations, The Effect of Preprocessing on Supervised Learning, Grid Search, Analyzing the result of cross-validation-Search over spaces that are not grids, Nested cross-validation, Algorithm Chains and Pipelines-Using Pipelines in Grid Searches, Convenient Pipeline Creation with make_pipeline-Grid-Searching Which Model To Use
- uncertainty estimates from, Uncertainty Estimates from Classifiers-Summary and Outlook
- cluster centers, k-Means Clustering
- clustering algorithms
- agglomerative clustering, Agglomerative Clustering-Hierarchical clustering and dendrograms
- applications for, Types of Unsupervised Learning
- comparing on faces dataset, Comparing algorithms on the faces dataset-Analyzing the faces dataset with agglomerative clustering
- DBSCAN, DBSCAN-DBSCAN
- evaluating with ground truth, Comparing and Evaluating Clustering Algorithms-Evaluating clustering without ground truth
- evaluating without ground truth, Evaluating clustering without ground truth-Evaluating clustering without ground truth
- goals of, Clustering
- k-means clustering, k-Means Clustering-Vector quantization, or seeing k-means as decomposition
- summary of, Summary of Clustering Methods
- code examples
- coef_ attribute, Linear regression (aka ordinary least squares), Ridge regression
- comments and questions, How to Contact Us
- competitions, Honing Your Skills
- conflation, Advanced Tokenization, Stemming, and Lemmatization
- confusion matrices, Confusion matrices-Precision, recall, and f-score
- context, Bag-of-Words with More Than One Word (n-Grams)
- continuous features, Representing Data and Engineering Features, Numbers Can Encode Categoricals
- core samples/core points, DBSCAN
- corpus, Types of Data Represented as Strings
- cos function, Univariate Nonlinear Transformations
- CountVectorizer, Bag-of-Words for Movie Reviews
- cross-validation
- analyzing results of, Analyzing the result of cross-validation-Analyzing the result of cross-validation
- benefits of, Benefits of Cross-Validation
- cross-validation splitters, More control over cross-validation
- grid search and, Grid Search with Cross-Validation
- in scikit-learn, Cross-Validation in scikit-learn
- leave-one-out cross-validation, Leave-one-out cross-validation
- nested, Nested cross-validation
- parallelizing with grid search, Parallelizing cross-validation and grid search
- principle of, Cross-Validation
- purpose of, Benefits of Cross-Validation
- shuffle-split cross-validation, Shuffle-split cross-validation
- stratified k-fold, Stratified k-Fold Cross-Validation and Other Strategies-Stratified k-Fold Cross-Validation and Other Strategies
- with groups, Cross-validation with groups
- cross_val_score function, Benefits of Cross-Validation, Parameter Selection with Preprocessing
D
- data points, defined, Problems Machine Learning Can Solve
- data representation, Representing Data and Engineering Features-Summary and Outlook (see also feature extraction/feature engineering; text data)
- automatic feature selection, Automatic Feature Selection-Iterative Feature Selection
- binning and, Binning, Discretization, Linear Models, and Trees-Binning, Discretization, Linear Models, and Trees
- categorical features, Categorical Variables-Binning, Discretization, Linear Models, and Trees
- effect on model performance, Representing Data and Engineering Features
- integer features, Numbers Can Encode Categoricals
- model complexity vs. dataset size, Relation of Model Complexity to Dataset Size
- overview of, Summary and Outlook
- table analogy, Problems Machine Learning Can Solve
- in training vs. test sets, Checking string-encoded categorical data
- understanding your data, Knowing Your Task and Knowing Your Data
- univariate nonlinear transformations, Univariate Nonlinear Transformations-Univariate Nonlinear Transformations
- data transformations, Applying Data Transformations
- data-driven research, Introduction
- DBSCAN
- decision boundaries, Analyzing KNeighborsClassifier, Linear models for classification
- decision function, The Decision Function
- decision trees
- analyzing, Analyzing decision trees
- building, Building decision trees
- controlling complexity of, Controlling complexity of decision trees
- data representation and, Binning, Discretization, Linear Models, and Trees-Binning, Discretization, Linear Models, and Trees
- feature importance in, Feature importance in trees
- if/else structure of, Decision Trees
- parameters, Strengths, weaknesses, and parameters
- vs. random forests, Random forests
- strengths and weaknesses, Strengths, weaknesses, and parameters
- decision_function, Taking uncertainty into account
- deep learning (see neural networks)
- dendrograms, Hierarchical clustering and dendrograms
- dense regions, DBSCAN
- dimensionality reduction, Principal Component Analysis (PCA), Non-Negative Matrix Factorization (NMF)
- discrete features, Representing Data and Engineering Features
- discretization, Binning, Discretization, Linear Models, and Trees-Binning, Discretization, Linear Models, and Trees
- distributed computing, Other Machine Learning Frameworks and Packages
- document clustering, Topic Modeling and Document Clustering
- documents, defined, Types of Data Represented as Strings
- dual_coef_ attribute, Understanding SVMs
F
- f(x)=y formula, Measuring Success: Training and Testing Data
- facial recognition, Eigenfaces for feature extraction, Applying NMF to face images
- factor analysis (FA), Applying NMF to face images
- false positive rate (FPR), Receiver operating characteristics (ROC) and AUC
- false positive/false negative errors, Kinds of errors
- feature extraction/feature engineering, Representing Data and Engineering Features-Summary and Outlook (see also data representation; text data)
- augmenting data with, Representing Data and Engineering Features
- automatic feature selection, Automatic Feature Selection-Iterative Feature Selection
- categorical features, Categorical Variables-Binning, Discretization, Linear Models, and Trees
- continuous vs. discrete features, Representing Data and Engineering Features
- defined, Problems Machine Learning Can Solve, Some Sample Datasets, Representing Data and Engineering Features
- interaction features, Interactions and Polynomials-Interactions and Polynomials
- with non-negative matrix factorization, Non-Negative Matrix Factorization (NMF)
- overview of, Summary and Outlook
- polynomial features, Interactions and Polynomials-Interactions and Polynomials
- with principal component analysis, Eigenfaces for feature extraction
- univariate nonlinear transformations, Univariate Nonlinear Transformations-Univariate Nonlinear Transformations
- using expert knowledge, Utilizing Expert Knowledge-Summary and Outlook
- feature importance, Feature importance in trees
- features, defined, Problems Machine Learning Can Solve
- feature_names attribute, Some Sample Datasets
- feed-forward neural networks, Neural Networks (Deep Learning)
- fit method, Building Your First Model: k-Nearest Neighbors, Strengths, weaknesses, and parameters, Estimating complexity in neural networks, Applying Data Transformations
- fit_transform method, Scaling Training and Test Data the Same Way
- floating-point numbers, Classification and Regression
- folds, Cross-Validation
- forge dataset, Some Sample Datasets
- frameworks, Other Machine Learning Frameworks and Packages
- free string data, Types of Data Represented as Strings
- freeform text data, Types of Data Represented as Strings
G
- gamma parameter, Tuning SVM parameters
- Gaussian kernels of SVC, The kernel trick, Tuning SVM parameters
- GaussianNB, Naive Bayes Classifiers
- generalization
- get_dummies function, Numbers Can Encode Categoricals
- get_support method of feature selection, Univariate Statistics
- gradient boosted regression trees
- for feature selection, Binning, Discretization, Linear Models, and Trees-Binning, Discretization, Linear Models, and Trees
- learning_rate parameter, Gradient boosted regression trees (gradient boosting machines)
- parameters, Strengths, weaknesses, and parameters
- vs. random forests, Gradient boosted regression trees (gradient boosting machines)
- strengths and weaknesses, Strengths, weaknesses, and parameters
- training set accuracy, Gradient boosted regression trees (gradient boosting machines)
- graphviz module, Analyzing decision trees
- grid search
- accessing pipeline attributes, Accessing Attributes in a Pipeline inside GridSearchCV
- alternate strategies for, Using different cross-validation strategies with grid search
- avoiding overfitting, The Danger of Overfitting the Parameters and the Validation Set
- model selection with, Grid-Searching Which Model To Use
- nested cross-validation, Nested cross-validation
- parallelizing with cross-validation, Parallelizing cross-validation and grid search
- pipeline preprocessing, Grid-Searching Preprocessing Steps and Model Parameters
- searching non-grid spaces, Search over spaces that are not grids
- simple example of, Simple Grid Search
- tuning parameters with, Grid Search
- using pipelines in, Using Pipelines in Grid Searches-Using Pipelines in Grid Searches
- with cross-validation, Grid Search with Cross-Validation
- GridSearchCV
H
- handcoded rules, disadvantages of, Why Machine Learning?
- heat maps, Applying PCA to the cancer dataset for visualization
- hidden layers, The neural network model
- hidden units, The neural network model
- hierarchical clustering, Hierarchical clustering and dendrograms
- high recall, Receiver operating characteristics (ROC) and AUC
- high-dimensional datasets, Some Sample Datasets
- histograms, Applying PCA to the cancer dataset for visualization
- hit rate, Precision, recall, and f-score
- hold-out sets, Measuring Success: Training and Testing Data
- human involvement/oversight, Humans in the Loop
I
- imbalanced datasets, Imbalanced datasets
- independent component analysis (ICA), Applying NMF to face images
- inference, Probabilistic Modeling, Inference, and Probabilistic Programming
- information leakage, Using Pipelines in Grid Searches
- information retrieval (IR), Types of Data Represented as Strings
- integer features, Numbers Can Encode Categoricals
- "intelligent" applications, Why Machine Learning?
- interactions, Some Sample Datasets, Interactions and Polynomials-Interactions and Polynomials
- intercept_ attribute, Linear regression (aka ordinary least squares)
- iris classification application
- iterative feature selection, Iterative Feature Selection
K
- k-fold cross-validation, Cross-Validation
- k-means clustering
- applying with scikit-learn, k-Means Clustering
- vs. classification, k-Means Clustering
- cluster centers, k-Means Clustering
- complex datasets, Vector quantization, or seeing k-means as decomposition
- evaluating and comparing, Evaluating clustering with ground truth
- example of, k-Means Clustering
- failures of, Failure cases of k-means
- strengths and weaknesses, Vector quantization, or seeing k-means as decomposition
- vector quantization with, Vector quantization, or seeing k-means as decomposition
- k-nearest neighbors (k-NN)
- Kaggle, Honing Your Skills
- kernelized support vector machines (SVMs)
- kernel trick, The kernel trick
- linear models and nonlinear features, Linear models and nonlinear features
- vs. linear support vector machines, Kernelized Support Vector Machines
- mathematics of, Kernelized Support Vector Machines
- parameters, Strengths, weaknesses, and parameters
- predictions with, Understanding SVMs
- preprocessing data for, Preprocessing data for SVMs
- strengths and weaknesses, Strengths, weaknesses, and parameters
- tuning SVM parameters, Tuning SVM parameters
- understanding, Understanding SVMs
- knn object, Building Your First Model: k-Nearest Neighbors
L
- L1 regularization, Lasso
- L2 regularization, Ridge regression, Linear models for classification, Strengths, weaknesses, and parameters
- Lasso model, Lasso
- Latent Dirichlet Allocation (LDA), Latent Dirichlet Allocation-Latent Dirichlet Allocation
- leafs, Decision Trees
- leakage, Using Pipelines in Grid Searches
- learn from the past approach, Utilizing Expert Knowledge
- learning_rate parameter, Gradient boosted regression trees (gradient boosting machines)
- leave-one-out cross-validation, Leave-one-out cross-validation
- lemmatization, Advanced Tokenization, Stemming, and Lemmatization-Advanced Tokenization, Stemming, and Lemmatization
- linear functions, Linear models for classification
- linear models
- classification, Linear models for classification
- data representation and, Binning, Discretization, Linear Models, and Trees-Binning, Discretization, Linear Models, and Trees
- vs. k-nearest neighbors, Linear models for regression
- Lasso, Lasso
- linear SVMs, Linear models for classification
- logistic regression, Linear models for classification
- multiclass classification, Linear models for multiclass classification
- ordinary least squares, Linear regression (aka ordinary least squares)
- parameters, Strengths, weaknesses, and parameters
- predictions with, Linear Models
- regression, Linear models for regression
- ridge regression, Ridge regression
- strengths and weaknesses, Strengths, weaknesses, and parameters
- linear regression, Linear regression (aka ordinary least squares), Interactions and Polynomials-Interactions and Polynomials
- linear support vector machines (SVMs), Linear models for classification
- linkage arrays, Hierarchical clustering and dendrograms
- live testing, Testing Production Systems
- log function, Univariate Nonlinear Transformations
- loss functions, Linear models for classification
- low-dimensional datasets, Some Sample Datasets
M
- machine learning
- algorithm chains and pipelines, Algorithm Chains and Pipelines-Summary and Outlook
- applications for, Why Machine Learning?-Knowing Your Task and Knowing Your Data
- approach to problem solving, Approaching a Machine Learning Problem-Honing Your Skills
- benefits of Python for, Why Python?
- building your own systems, Preface
- data representation, Representing Data and Engineering Features-Summary and Outlook
- examples of, Introduction, A First Application: Classifying Iris Species-Evaluating the Model
- mathematics of, Who Should Read This Book
- model evaluation and improvement, Model Evaluation and Improvement-Summary and Outlook
- preprocessing and scaling, Preprocessing and Scaling-The Effect of Preprocessing on Supervised Learning
- prerequisites to learning, Who Should Read This Book
- resources, Online Resources, Where to Go from Here-Honing Your Skills
- scikit-learn and, scikit-learn-Versions Used in this Book
- supervised learning, Supervised Learning-Summary and Outlook
- understanding your data, Knowing Your Task and Knowing Your Data
- unsupervised learning, Unsupervised Learning and Preprocessing-Summary and Outlook
- working with text data, Working with Text Data-Summary and Outlook
- make_pipeline function
- manifold learning algorithms
- mathematical functions for feature transformations, Univariate Nonlinear Transformations
- matplotlib, matplotlib
- max_features parameter, Building random forests
- meta-estimators for trees and forests, Grid Search with Cross-Validation
- method chaining, Strengths, weaknesses, and parameters
- metrics (see evaluation metrics and scoring)
- mglearn, mglearn
- mllib, Other Machine Learning Frameworks and Packages
- model-based feature selection, Model-Based Feature Selection
- models (see also algorithms)
- calibrated, Taking uncertainty into account
- capable of generalization, Generalization, Overfitting, and Underfitting
- coefficients with text data, Investigating Model Coefficients-Advanced Tokenization, Stemming, and Lemmatization
- complexity vs. dataset size, Relation of Model Complexity to Dataset Size
- cross-validation of, Cross-Validation-Cross-validation with groups
- effect of data representation choices on, Representing Data and Engineering Features
- evaluation and improvement, Model Evaluation and Improvement-Model Evaluation and Improvement
- evaluation metrics and scoring, Evaluation Metrics and Scoring-Summary and Outlook
- iris classification application, A First Application: Classifying Iris Species-Evaluating the Model
- overfitting vs. underfitting, Generalization, Overfitting, and Underfitting
- pipeline preprocessing and, Grid-Searching Preprocessing Steps and Model Parameters
- selecting, Using Evaluation Metrics in Model Selection
- selecting with grid search, Grid-Searching Which Model To Use
- theory behind, Theory
- tuning parameters with grid search, Grid Search
- movie reviews, Example Application: Sentiment Analysis of Movie Reviews
- multiclass classification
- multilayer perceptrons (MLPs), Neural Networks (Deep Learning)
- MultinomialNB, Naive Bayes Classifiers
N
- n-grams, Bag-of-Words with More Than One Word (n-Grams)
- naive Bayes classifiers
- natural language processing (NLP), Types of Data Represented as Strings, Summary and Outlook
- negative class, Classification and Regression
- nested cross-validation, Nested cross-validation
- Netflix prize challenge, Ranking, Recommender Systems, and Other Kinds of Learning
- neural networks (deep learning)
- non-negative matrix factorization (NMF)
- normalization, Advanced Tokenization, Stemming, and Lemmatization
- normalized mutual information (NMI), Evaluating clustering with ground truth
- NumPy (Numeric Python) library, NumPy
O
- offline evaluation, Testing Production Systems
- one-hot-encoding, One-Hot-Encoding (Dummy Variables)-Checking string-encoded categorical data
- one-out-of-N encoding, One-Hot-Encoding (Dummy Variables)-Checking string-encoded categorical data
- one-vs.-rest approach, Linear models for multiclass classification
- online resources, Online Resources
- online testing, Testing Production Systems
- OpenML platform, Honing Your Skills
- operating points, Precision-recall curves and ROC curves
- ordinary least squares (OLS), Linear models for regression
- out-of-core learning, Scaling to Larger Datasets
- outlier detection, Analyzing the faces dataset with DBSCAN
- overfitting, Generalization, Overfitting, and Underfitting, The Danger of Overfitting the Parameters and the Validation Set
P
- pair plots, First Things First: Look at Your Data
- pandas
- parallelization over a cluster, Scaling to Larger Datasets
- permissions, Using Code Examples
- pipelines (see algorithm chains and pipelines)
- polynomial features, Interactions and Polynomials-Interactions and Polynomials
- polynomial kernels, The kernel trick
- polynomial regression, Interactions and Polynomials
- positive class, Classification and Regression
- POSIX time, Utilizing Expert Knowledge
- pre- and post-pruning, Controlling complexity of decision trees
- precision, Precision, recall, and f-score, Humans in the Loop
- precision-recall curves, Precision-recall curves and ROC curves-Precision-recall curves and ROC curves
- predict for the future approach, Utilizing Expert Knowledge
- predict method, Making Predictions, k-Neighbors classification, Strengths, weaknesses, and parameters, Grid Search with Cross-Validation
- predict_proba function, Predicting Probabilities, Taking uncertainty into account
- preprocessing, Preprocessing and Scaling-The Effect of Preprocessing on Supervised Learning
- principal component analysis (PCA)
- probabilistic modeling, Probabilistic Modeling, Inference, and Probabilistic Programming
- probabilistic programming, Probabilistic Modeling, Inference, and Probabilistic Programming
- problem solving
- production systems
- pruning for decision trees, Controlling complexity of decision trees
- pseudorandom number generators, Measuring Success: Training and Testing Data
- pure leafs, Building decision trees
- PyMC language, Probabilistic Modeling, Inference, and Probabilistic Programming
- Python
R
- R language, Other Machine Learning Frameworks and Packages
- radial basis function (RBF) kernel, The kernel trick
- random forests
- analyzing, Analyzing random forests
- building, Building random forests
- data representation and, Binning, Discretization, Linear Models, and Trees-Binning, Discretization, Linear Models, and Trees
- vs. decision trees, Random forests
- vs. gradient boosted regression trees, Gradient boosted regression trees (gradient boosting machines)
- parameters, Strengths, weaknesses, and parameters
- predictions with, Building random forests
- randomization in, Random forests
- strengths and weaknesses, Strengths, weaknesses, and parameters
- random_state parameter, Measuring Success: Training and Testing Data
- ranking, Ranking, Recommender Systems, and Other Kinds of Learning
- real numbers, Classification and Regression
- recall, Precision, recall, and f-score
- receiver operating characteristics (ROC) curves, Receiver operating characteristics (ROC) and AUC-Receiver operating characteristics (ROC) and AUC
- recommender systems, Ranking, Recommender Systems, and Other Kinds of Learning
- rectified linear unit (relu), The neural network model
- rectifying nonlinearity, The neural network model
- recurrent neural networks (RNNs), Summary and Outlook
- recursive feature elimination (RFE), Iterative Feature Selection
- regression
- regression problems
- Boston Housing dataset, Some Sample Datasets
- vs. classification problems, Classification and Regression
- evaluation metrics and scoring, Regression Metrics
- examples of, Classification and Regression
- goals for, Classification and Regression
- k-nearest neighbors, k-neighbors regression
- Lasso, Lasso
- linear models, Linear models for regression
- ridge regression, Ridge regression
- wave dataset illustration, Some Sample Datasets
- regularization
- rescaling
- resources, Online Resources
- ridge regression, Ridge regression
- robustness-based clustering, Evaluating clustering without ground truth
- roots, Building decision trees
S
- samples, defined, Problems Machine Learning Can Solve
- scaling, Preprocessing and Scaling-The Effect of Preprocessing on Supervised Learning
- scatter plots, First Things First: Look at Your Data
- scikit-learn
- alternate frameworks, Other Machine Learning Frameworks and Packages
- benefits of, scikit-learn
- Bunch objects, Some Sample Datasets
- cancer dataset, Some Sample Datasets
- core code for, Summary and Outlook
- data and labels in, Measuring Success: Training and Testing Data
- documentation, scikit-learn
- feature_names attribute, Some Sample Datasets
- fit method, Building Your First Model: k-Nearest Neighbors, Strengths, weaknesses, and parameters, Estimating complexity in neural networks, Applying Data Transformations
- fit_transform method, Scaling Training and Test Data the Same Way
- installing, Installing scikit-learn
- knn object, Building Your First Model: k-Nearest Neighbors
- libraries and tools, Essential Libraries and Tools-mglearn
- predict method, Making Predictions, k-Neighbors classification, Strengths, weaknesses, and parameters
- Python 2 vs. Python 3, Python 2 Versus Python 3
- random_state parameter, Measuring Success: Training and Testing Data
- scaling mechanisms in, The Effect of Preprocessing on Supervised Learning
- score method, Evaluating the Model, k-Neighbors classification, k-neighbors regression
- transform method, Applying Data Transformations
- user guide, scikit-learn
- versions used, Versions Used in this Book
- scikit-learn classes and functions
- accuracy_score, Evaluating clustering with ground truth
- adjusted_rand_score, Evaluating clustering with ground truth
- AgglomerativeClustering, Agglomerative Clustering, Evaluating clustering with ground truth, Analyzing the faces dataset with agglomerative clustering-Analyzing the faces dataset with agglomerative clustering
- average_precision_score, Precision-recall curves and ROC curves
- BaseEstimator, Building Your Own Estimator
- classification_report, Precision, recall, and f-score-Taking uncertainty into account, Metrics for Multiclass Classification
- confusion_matrix, Confusion matrices-Regression Metrics
- CountVectorizer, Applying Bag-of-Words to a Toy Dataset-Summary and Outlook
- cross_val_score, Cross-Validation in scikit-learn, More control over cross-validation, Using Evaluation Metrics in Model Selection, Parameter Selection with Preprocessing, Building Your Own Estimator
- DBSCAN, DBSCAN-DBSCAN
- DecisionTreeClassifier, Controlling complexity of decision trees, Imbalanced datasets
- DecisionTreeRegressor, Controlling complexity of decision trees, Feature importance in trees
- DummyClassifier, Imbalanced datasets
- ElasticNet class, Lasso
- ENGLISH_STOP_WORDS, Stopwords
- Estimator, Building Your First Model: k-Nearest Neighbors
- export_graphviz, Analyzing decision trees
- f1_score, Precision, recall, and f-score, Precision-recall curves and ROC curves
- fetch_lfw_people, Eigenfaces for feature extraction
- f_regression, Univariate Statistics, Using Pipelines in Grid Searches
- GradientBoostingClassifier, Gradient boosted regression trees (gradient boosting machines)-Gradient boosted regression trees (gradient boosting machines), Uncertainty Estimates from Classifiers, Uncertainty in Multiclass Classification
- GridSearchCV, Grid Search with Cross-Validation, Using Evaluation Metrics in Model Selection-Using Evaluation Metrics in Model Selection, Algorithm Chains and Pipelines-Using Pipelines in Grid Searches, Accessing Attributes in a Pipeline inside GridSearchCV-Grid-Searching Which Model To Use, Building Your Own Estimator
- GroupKFold, Cross-validation with groups
- KFold, More control over cross-validation, Cross-validation with groups
- KMeans, Failure cases of k-means-Vector quantization, or seeing k-means as decomposition
- KNeighborsClassifier, Building Your First Model: k-Nearest Neighbors-Summary and Outlook, k-Neighbors classification-k-neighbors regression
- KNeighborsRegressor, k-neighbors regression-Linear models for regression
- Lasso, Lasso-Lasso
- LatentDirichletAllocation, Latent Dirichlet Allocation
- LeaveOneOut, Leave-one-out cross-validation
- LinearRegression, Linear regression (aka ordinary least squares)-Linear models for classification, Feature importance in trees, Utilizing Expert Knowledge
- LinearSVC, Linear models for classification-Linear models for classification, Linear models for multiclass classification, Strengths, weaknesses, and parameters, Naive Bayes Classifiers
- load_boston, Some Sample Datasets, Interactions and Polynomials, Grid-Searching Preprocessing Steps and Model Parameters
- load_breast_cancer, Some Sample Datasets, Analyzing KNeighborsClassifier, Linear models for classification, Controlling complexity of decision trees, Applying Data Transformations, Applying PCA to the cancer dataset for visualization, Univariate Statistics, Algorithm Chains and Pipelines
- load_digits, Manifold Learning with t-SNE, Imbalanced datasets
- load_iris, Meet the Data, Uncertainty in Multiclass Classification, Cross-Validation in scikit-learn
- LogisticRegression, Linear models for classification-Linear models for classification, Strengths, weaknesses, and parameters, Summary and Outlook, Cross-Validation in scikit-learn, Imbalanced datasets, Accessing Attributes in a Pipeline inside GridSearchCV, Bag-of-Words for Movie Reviews-Advanced Tokenization, Stemming, and Lemmatization
- make_blobs, Linear models and nonlinear features, Uncertainty Estimates from Classifiers, Scaling Training and Test Data the Same Way, Failure cases of k-means-Agglomerative Clustering, DBSCAN, Taking uncertainty into account
- make_circles, Uncertainty Estimates from Classifiers
- make_moons, Analyzing random forests, Tuning neural networks, Failure cases of k-means, DBSCAN-Evaluating clustering without ground truth
- make_pipeline, Convenient Pipeline Creation with make_pipeline-Grid-Searching Preprocessing Steps and Model Parameters
- MinMaxScaler, Preprocessing data for SVMs, Different Kinds of Preprocessing, Applying Data Transformations-The Effect of Preprocessing on Supervised Learning, DBSCAN, Interactions and Polynomials, Building Pipelines, Using Pipelines in Grid Searches, Grid-Searching Which Model To Use
- MLPClassifier, The neural network model-Estimating complexity in neural networks
- NMF, Dimensionality Reduction, Feature Extraction, and Manifold Learning, Applying NMF to face images-Applying NMF to face images, Vector quantization, or seeing k-means as decomposition-Agglomerative Clustering, Latent Dirichlet Allocation
- Normalizer, Different Kinds of Preprocessing
- OneHotEncoder, Numbers Can Encode Categoricals, Utilizing Expert Knowledge
- ParameterGrid, Nested cross-validation
- PCA, Principal Component Analysis (PCA)-Manifold Learning with t-SNE, Vector quantization, or seeing k-means as decomposition, Comparing algorithms on the faces dataset-Analyzing the faces dataset with agglomerative clustering, The General Pipeline Interface-Accessing Step Attributes, Latent Dirichlet Allocation
- Pipeline, Algorithm Chains and Pipelines-Grid-Searching Which Model To Use, Summary and Outlook
- PolynomialFeatures, Interactions and Polynomials-Interactions and Polynomials, Utilizing Expert Knowledge, Grid-Searching Preprocessing Steps and Model Parameters
- precision_recall_curve, Precision-recall curves and ROC curves-Precision-recall curves and ROC curves
- RandomForestClassifier, Building random forests-Analyzing random forests, Model-Based Feature Selection, Precision-recall curves and ROC curves, Grid-Searching Which Model To Use
- RandomForestRegressor, Building random forests, Interactions and Polynomials, Model-Based Feature Selection
- RFE, Iterative Feature Selection-Iterative Feature Selection
- Ridge, Ridge regression, Strengths, weaknesses, and parameters, Tuning neural networks, Interactions and Polynomials, Univariate Nonlinear Transformations, Using Pipelines in Grid Searches, Grid-Searching Preprocessing Steps and Model Parameters-Grid-Searching Preprocessing Steps and Model Parameters
- RobustScaler, Different Kinds of Preprocessing
- roc_auc_score, Receiver operating characteristics (ROC) and AUC-Using Evaluation Metrics in Model Selection
- roc_curve, Receiver operating characteristics (ROC) and AUC-Receiver operating characteristics (ROC) and AUC
- SCORERS, Using Evaluation Metrics in Model Selection
- SelectFromModel, Model-Based Feature Selection
- SelectPercentile, Univariate Statistics, Using Pipelines in Grid Searches
- ShuffleSplit, Shuffle-split cross-validation, Shuffle-split cross-validation
- silhouette_score, Evaluating clustering without ground truth
- StandardScaler, Tuning neural networks, Different Kinds of Preprocessing, Scaling Training and Test Data the Same Way, Applying PCA to the cancer dataset for visualization, Eigenfaces for feature extraction, DBSCAN-Evaluating clustering without ground truth, Convenient Pipeline Creation with make_pipeline-Grid-Searching Which Model To Use
- StratifiedKFold, Cross-validation with groups, Nested cross-validation
- StratifiedShuffleSplit, Shuffle-split cross-validation, Advanced Tokenization, Stemming, and Lemmatization
- SVC, Linear models for classification, Tuning SVM parameters, Applying Data Transformations, The Effect of Preprocessing on Supervised Learning, Grid Search-Grid Search with Cross-Validation, Analyzing the result of cross-validation-Search over spaces that are not grids, Algorithm Chains and Pipelines-Using Pipelines in Grid Searches, Convenient Pipeline Creation with make_pipeline-Grid-Searching Which Model To Use
- SVR, Kernelized Support Vector Machines, Interactions and Polynomials
- TfidfVectorizer, Rescaling the Data with tf–idf-Summary and Outlook
- train_test_split, Measuring Success: Training and Testing Data-First Things First: Look at Your Data, Model Evaluation and Improvement, Taking uncertainty into account, Precision-recall curves and ROC curves
- TransformerMixin, Testing Production Systems
- TSNE, Manifold Learning with t-SNE
- SciPy, SciPy
- score method, Evaluating the Model, k-Neighbors classification, k-neighbors regression, Grid Search with Cross-Validation, Building Pipelines
- sensitivity, Precision, recall, and f-score
- sentiment analysis example, Example Application: Sentiment Analysis of Movie Reviews
- shapes, defined, Meet the Data
- shuffle-split cross-validation, Shuffle-split cross-validation
- sin function, Univariate Nonlinear Transformations
- soft voting strategy, Building random forests
- spark computing environment, Other Machine Learning Frameworks and Packages
- sparse coding (dictionary learning), Applying NMF to face images
- splits, Cross-Validation
- Stan language, Probabilistic Modeling, Inference, and Probabilistic Programming
- statsmodel package, Other Machine Learning Frameworks and Packages
- stemming, Advanced Tokenization, Stemming, and Lemmatization-Advanced Tokenization, Stemming, and Lemmatization
- stopwords, Stopwords
- stratified k-fold cross-validation, Stratified k-Fold Cross-Validation and Other Strategies-Stratified k-Fold Cross-Validation and Other Strategies
- string-encoded categorical data, Checking string-encoded categorical data
- supervised learning, Supervised Learning-Summary and Outlook (see also classification problems; regression problems)
- algorithms for
- decision trees, Decision Trees-Strengths, weaknesses, and parameters
- ensembles of decision trees, Ensembles of Decision Trees-Strengths, weaknesses, and parameters
- k-nearest neighbors, k-Nearest Neighbors-Strengths, weaknesses, and parameters
- kernelized support vector machines, Kernelized Support Vector Machines-Strengths, weaknesses, and parameters
- linear models, Linear Models-Strengths, weaknesses, and parameters
- naive Bayes classifiers, Naive Bayes Classifiers
- neural networks (deep learning), Neural Networks (Deep Learning)-Estimating complexity in neural networks
- overview of, Problems Machine Learning Can Solve
- data representation, Problems Machine Learning Can Solve
- examples of, Problems Machine Learning Can Solve
- generalization, Generalization, Overfitting, and Underfitting
- goals for, Supervised Learning
- model complexity vs. dataset size, Relation of Model Complexity to Dataset Size
- overfitting vs. underfitting, Generalization, Overfitting, and Underfitting
- overview of, Summary and Outlook
- sample datasets, Some Sample Datasets-Some Sample Datasets
- uncertainty estimates, Uncertainty Estimates from Classifiers-Summary and Outlook
- support vectors, Understanding SVMs
- synthetic datasets, Some Sample Datasets
T
- t-SNE algorithm (see manifold learning algorithms)
- tangens hyperbolicus (tanh), The neural network model
- term frequency–inverse document frequency (tf–idf), Rescaling the Data with tf–idf-Advanced Tokenization, Stemming, and Lemmatization
- terminal nodes, Decision Trees
- test data/test sets
- text data, Working with Text Data-Summary and Outlook
- time series predictions, Ranking, Recommender Systems, and Other Kinds of Learning
- tokenization, Representing Text Data as a Bag of Words, Advanced Tokenization, Stemming, and Lemmatization-Advanced Tokenization, Stemming, and Lemmatization
- top nodes, Building decision trees
- topic modeling, with LDA, Topic Modeling and Document Clustering-Latent Dirichlet Allocation
- training data, Measuring Success: Training and Testing Data
- train_test_split function, Benefits of Cross-Validation
- transform method, Applying Data Transformations, The General Pipeline Interface, Bag-of-Words for Movie Reviews
- transformations
- tree module, Analyzing decision trees
- trigrams, Bag-of-Words with More Than One Word (n-Grams)
- true positive rate (TPR), Precision, recall, and f-score, Receiver operating characteristics (ROC) and AUC
- true positives/true negatives, Confusion matrices
- typographical conventions, Conventions Used in This Book
..................Content has been hidden....................
You can't read the all page of ebook, please click
here login for view all page.