5. Postmodel Workflow
by Gavin Hackeling, Trent Hauck, Guillermo Moncecchi, Raúl Garreta
scikit-learn : Machine Learning Simplified
- scikit-learn: Machine Learning Simplified
- Table of Contents
- Credits
- Preface
- 1. Module 1
- 2. Module 2
- 1. Premodel Workflow
- Introduction
- Getting sample data from external sources
- Creating sample data for toy analysis
- Scaling data to the standard normal
- Creating binary features through thresholding
- Working with categorical variables
- Binarizing label features
- Imputing missing values through various strategies
- Using Pipelines for multiple preprocessing steps
- Reducing dimensionality with PCA
- Using factor analysis for decomposition
- Kernel PCA for nonlinear dimensionality reduction
- Using truncated SVD to reduce dimensionality
- Decomposition to classify with DictionaryLearning
- Putting it all together with Pipelines
- Using Gaussian processes for regression
- Defining the Gaussian process object directly
- Using stochastic gradient descent for regression
- 2. Working with Linear Models
- Introduction
- Fitting a line through data
- Evaluating the linear regression model
- Using ridge regression to overcome linear regression's shortfalls
- Optimizing the ridge regression parameter
- Using sparsity to regularize models
- Taking a more fundamental approach to regularization with LARS
- Using linear methods for classification – logistic regression
- Directly applying Bayesian ridge regression
- Using boosting to learn from errors
- 3. Building Models with Distance Metrics
- Introduction
- Using KMeans to cluster data
- Optimizing the number of centroids
- Assessing cluster correctness
- Using MiniBatch KMeans to handle more data
- Quantizing an image with KMeans clustering
- Finding the closest objects in the feature space
- Probabilistic clustering with Gaussian Mixture Models
- Using KMeans for outlier detection
- Using k-NN for regression
- 4. Classifying Data with scikit-learn
- Introduction
- Doing basic classifications with Decision Trees
- Tuning a Decision Tree model
- Using many Decision Trees – random forests
- Tuning a random forest model
- Classifying data with support vector machines
- Generalizing with multiclass classification
- Using LDA for classification
- Working with QDA – a nonlinear LDA
- Using Stochastic Gradient Descent for classification
- Classifying documents with Naïve Bayes
- Label propagation with semi-supervised learning
- 5. Postmodel Workflow
- Introduction
- K-fold cross validation
- Automatic cross validation
- Cross validation with ShuffleSplit
- Stratified k-fold
- Poor man's grid search
- Brute force grid search
- Using dummy estimators to compare results
- Regression model evaluation
- Feature selection
- Feature selection on L1 norms
- Persisting models with joblib
- 1. Premodel Workflow
- 3. Module 3
- 1. The Fundamentals of Machine Learning
- 2. Linear Regression
- 3. Feature Extraction and Preprocessing
- 4. From Linear Regression to Logistic Regression
- 5. Nonlinear Classification and Regression with Decision Trees
- 6. Clustering with K-Means
- 7. Dimensionality Reduction with PCA
- 8. The Perceptron
- 9. From the Perceptron to Support Vector Machines
- 10. From the Perceptron to Artificial Neural Networks
- Bibliography
- Index
This chapter will cover the following recipes:
- K-fold cross validation
- Automatic cross validation
- Cross validation with ShuffleSplit
- Stratified k-fold
- Poor man's grid search
- Brute force grid search
- Using dummy estimators to compare results
- Regression model evaluation
- Feature selection
- Feature selection on L1 norms
- Persisting models with joblib
Even though by design the chapters are unordered, you could argue by virtue of the art of data science, we've saved the best for last.
For the most part, each recipe within this chapter is applicable to the various models we've worked with. In some ways, you can think about this chapter as tuning the parameters and features. Ultimately, we need to choose some criteria to determine the "best" model. We'll use various measures to define best. This is covered in the Regression model evaluation recipe. Then in the Cross validation with ShuffleSplit recipe, we will randomize the evaluation across subsets of the data to help avoid overfitting.
-
No Comment
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.