Feature Engineering Made Easy by Ozdemir, Sinan -- Read -- Imperial Library of Trantor

Index

Title Page Copyright and Credits

Feature Engineering Made Easy

Packt Upsell

Why subscribe? PacktPub.com

Contributors

About the authors About the reviewer Packt is searching for authors like you

Preface

Who this book is for What this book covers To get the most out of this book

Download the example code files Download the color images Conventions used

Get in touch

Reviews

Introduction to Feature Engineering

Motivating example – AI-powered communications Why feature engineering matters What is feature engineering?

Understanding the basics of data and machine learning

Supervised learning Unsupervised learning

Unsupervised learning example – marketing segments

Evaluation of machine learning algorithms and feature engineering procedures

Example of feature engineering procedures – can anyone really predict the weather? Steps to evaluate a feature engineering procedure Evaluating supervised learning algorithms Evaluating unsupervised learning algorithms

Feature understanding – what’s in my dataset? Feature improvement – cleaning datasets Feature selection – say no to bad attributes Feature construction – can we build it? Feature transformation – enter math-man Feature learning – using AI to better our AI Summary

Feature Understanding – What's in My Dataset?

The structure, or lack thereof, of data An example of unstructured data – server logs Quantitative versus qualitative data

Salary ranges by job classification

The four levels of data

The nominal level

Mathematical operations allowed

The ordinal level

Mathematical operations allowed

The interval level

Mathematical operations allowed Plotting two columns at the interval level

The ratio level

Mathematical operations allowed

Recap of the levels of data Summary

Feature Improvement - Cleaning Datasets

Identifying missing values in data

The Pima Indian Diabetes Prediction dataset The exploratory data analysis (EDA)

Dealing with missing values in a dataset

Removing harmful rows of data Imputing the missing values in data Imputing values in a machine learning pipeline

Pipelines in machine learning

Standardization and normalization

Z-score standardization The min-max scaling method The row normalization method Putting it all together

Summary

Feature Construction

Examining our dataset Imputing categorical features

Custom imputers Custom category imputer Custom quantitative imputer

Encoding categorical variables

Encoding at the nominal level Encoding at the ordinal level Bucketing continuous features into categories Creating our pipeline

Extending numerical features

Activity recognition from the Single Chest-Mounted Accelerometer dataset Polynomial features

Parameters Exploratory data analysis

Text-specific feature construction

Bag of words representation CountVectorizer

CountVectorizer parameters

The Tf-idf vectorizer Using text in machine learning pipelines

Summary

Feature Selection

Achieving better performance in feature engineering

A case study – a credit card defaulting dataset

Creating a baseline machine learning pipeline The types of feature selection

Statistical-based feature selection

Using Pearson correlation to select features Feature selection using hypothesis testing

Interpreting the p-value Ranking the p-value

Model-based feature selection

A brief refresher on natural language processing Using machine learning to select features

Tree-based model feature selection metrics

Linear models and regularization

A brief introduction to regularization Linear model coefficients as another feature importance metric

Choosing the right feature selection method Summary

Feature Transformations

Dimension reduction – feature transformations versus feature selection versus feature construction Principal Component Analysis

How PCA works PCA with the Iris dataset – manual example

Creating the covariance matrix of the dataset Calculating the eigenvalues of the covariance matrix Keeping the top k eigenvalues (sorted by the descending eigenvalues) Using the kept eigenvectors to transform new data-points

Scikit-learn's PCA How centering and scaling data affects PCA A deeper look into the principal components Linear Discriminant Analysis

How LDA works

Calculating the mean vectors of each class Calculating within-class and between-class scatter matrices Calculating eigenvalues and eigenvectors for SW-1SB Keeping the top k eigenvectors by ordering them by descending eigenvalues Using the top eigenvectors to project onto the new space

How to use LDA in scikit-learn

LDA versus PCA – iris dataset Summary

Feature Learning

Parametric assumptions of data

Non-parametric fallacy The algorithms of this chapter

Restricted Boltzmann Machines

Not necessarily dimension reduction The graph of a Restricted Boltzmann Machine The restriction of a Boltzmann Machine Reconstructing the data MNIST dataset

The BernoulliRBM

Extracting PCA components from MNIST

Extracting RBM components from MNIST Using RBMs in a machine learning pipeline

Using a linear model on raw pixel values Using a linear model on extracted PCA components Using a linear model on extracted RBM components

Learning text features – word vectorizations

Word embeddings Two approaches to word embeddings - Word2vec and GloVe Word2Vec - another shallow neural network The gensim package for creating Word2vec embeddings Application of word embeddings - information retrieval

Summary

Case Studies

Case study 1 - facial recognition

Applications of facial recognition The data Some data exploration Applied facial recognition

Case study 2 - predicting topics of hotel reviews data

Applications of text clustering Hotel review data Exploration of the data The clustering model SVD versus PCA components Latent semantic analysis

Summary

Other Books You May Enjoy

Leave a review - let other readers know what you think

← Prev
Back
Next →

← Prev
Back
Next →