Predictive Analytics with TensorFlow电子书

Predictive Analytics with TensorFlow

Predictive Analytics with TensorFlow

Credits

About the Author

Acknowledgments

About the Reviewers

www.PacktPub.com

eBooks, discount offers, and more

Why subscribe?

Customer Feedback

Preface

What this book covers

What you need for this book

Who this book is for

Conventions

Reader feedback

Customer support

Downloading the example code

Downloading the color images of this book

Errata

Piracy

Questions

1. Basic Python and Linear Algebra for Predictive Analytics

A basic introduction to predictive analytics

Why predictive analytics?

Working principles of a predictive model

A bit of linear algebra

Programming linear algebra

Installing and getting started with Python

Installing on Windows

Installing Python on Linux

Installing and upgrading PIP (or PIP3)

Installing Python on Mac OS

Installing packages in Python

Getting started with Python

Python data types

Using strings in Python

Using lists in Python

Using tuples in Python

Using dictionary in Python

Using sets in Python

Functions in Python

Classes in Python

Vectors, matrices, and graphs

Vectors

Matrices

Matrix addition

Matrix subtraction

Multiplying two matrices

Finding the determinant of a matrix

Finding the transpose of a matrix

Solving simultaneous linear equations

Eigenvalues and eigenvectors

Span and linear independence

Principal component analysis

Singular value decomposition

Data compression in a predictive model using SVD

Predictive analytics tools in Python

Summary

2. Statistics, Probability, and Information Theory for Predictive Modeling

Using statistics in predictive modeling

Statistical models

Parametric versus nonparametric model

Parametric predictive models

Nonparametric predictive models

Population and sample

Random sampling

Expectation

Central limit theorem

Skewness and data distribution

Standard deviation and variance

Covariance and correlation

Interquartile, range, and quartiles

Hypothesis testing

Chi-square tests

Chi-square independence test

Basic probability for predictive modeling

Probability and the random variables

Generating random numbers and setting the seed

Probability distributions

Marginal probability

Conditional probability

The chain rule of conditional probability

Independence and conditional independence

Bayes' rule

Using information theory in predictive modeling

Self-information

Mutual information

Entropy

Shannon entropy

Joint entropy

Conditional entropy

Information gain

Using information theory

Using information theory in Python

Summary

3. From Data to Decisions – Getting Started with TensorFlow

Taking decisions based on data - Titanic example

Data value chain for making decisions

From disaster to decision – Titanic survival example

General overview of TensorFlow

Installing and configuring TensorFlow

Installing TensorFlow on Linux

Installing Python and nVidia driver

Installing NVIDIA CUDA

Installing NVIDIA cuDNN v5.1+

Installing the libcupti-dev library

Installing TensorFlow

Installing TensorFlow with native pip

Installing with virtualenv

Installing TensorFlow from source

Testing your TensorFlow installation

TensorFlow computational graph

TensorFlow programming model

Data model in TensorFlow

Tensors

Rank

Shape

Data type

Variables

Fetches

Feeds and placeholders

TensorBoard

How does TensorBoard work?

Getting started with TensorFlow – linear regression and beyond

Source code for the linear regression

Summary

4. Putting Data in Place - Supervised Learning for Predictive Analytics

Supervised learning for predictive analytics

Linear regression - revisited

Problem statement

Using linear regression for movie rating prediction

From disaster to decision - Titanic example revisited

An exploratory analysis of the Titanic dataset

Feature engineering

Logistic regression for survival prediction

Using TensorFlow contrib

Linear SVM for survival prediction

Ensemble method for survival prediction: random forest

A comparative analysis

Summary

5. Clustering Your Data - Unsupervised Learning for Predictive Analytics

Unsupervised learning and clustering

Using K-means for predictive analytics

How K-means works

Using K-means for predicting neighborhoods

Predictive models for clustering audio files

Using kNN for predictive analytics

Working principles of kNN

Implementing a kNN-based predictive model

Summary

6. Predictive Analytics Pipelines for NLP

NLP analytics pipelines

Using text analytics

Transformers and estimators

Standard transformer

Estimator transformer

StopWordsRemover

N-gram

Using BOW for predictive analytics

Bag-of-words

The problem definition

The dataset description and exploration

Spam prediction using LR and BOW with TensorFlow

TF-IDF model for predictive analytics

How to compute TF, IDF, and TFIDF?

Implementing a TF-IDF model for spam prediction

Using Word2vec for sentiment analysis

Continuous bag-of-words

Continuous skip-gram

Using CBOW for word embedding and model building

CBOW model building

Reusing the CBOW for predicting sentiment

Summary

7. Using Deep Neural Networks for Predictive Analytics

Deep learning for better predictive analytics

Artificial Neural Networks

Deep Neural Networks

DNN architectures

Multilayer perceptrons

Training an MLP

Using MLPs

DNN performance analysis

Fine-tuning DNN hyperparameters

Number of hidden layers

Number of neurons per hidden layer

Activation functions

Weight and biases initialization

Regularization

Using multilayer perceptrons for predictive analytics

Dataset description

Preprocessing

A TensorFlow implementation of MLP

Deep belief networks

Restricted Boltzmann Machines

Construction of a simple DBN

Unsupervised Pretraining

Using deep belief networks for predictive analytics

Summary

8. Using Convolutional Neural Networks for Predictive Analytics

CNNs and the drawbacks of regular DNNs

CNN architecture

Convolutional operations

Applying convolution operations in TensorFlow

Pooling layer and padding operations

Applying subsampling operations in TensorFlow

Tuning CNN hyperparameters

CNN-based predictive model for sentiment analysis

Exploring movie and product review datasets

Using CNN for predictive analytics about movie reviews

CNN model for emotion recognition

Dataset description

CNN architecture design

Testing the model on your own image

Using complex CNN for predictive analytics

Dataset description

CNN predictive model for image classification

Summary

9. Using Recurrent Neural Networks for Predictive Analytics

RNN architecture

Contextual information and the architecture of RNNs

BRNNs

LSTM networks

GRU cell

Using BRNN for image classification

Implementing an RNN for spam prediction

Developing a predictive model for time series data

Description of the dataset

Preprocessing and exploratory analysis

LSTM predictive model

Model evaluation

An LSTM predictive model for sentiment analysis

Network design

LSTM model training

Visualizing through TensorBoard

LSTM model evaluation

Summary

10. Recommendation Systems for Predictive Analytics

Recommendation systems

Collaborative filtering approaches

Content-based filtering approaches

Hybrid recommendation systems

Model-based collaborative filtering

Collaborative filtering approach for movie recommendations

The utility matrix

Dataset description

Ratings data

Movies data

User data

Exploratory analysis of the dataset

Implementing a movie recommendation engine

Training the model with available ratings

Inferencing the saved model

Generating a user-item table

Clustering similar movies

Movie rating prediction by users

Finding the top K movies

Predicting top K similar movies

Computing the user-user similarity

Evaluating the recommendation system

Factorization machines for recommendation systems

Factorization machines

The cold start problem in recommendation systems

Problem definition and formulation

Dataset description

Preprocessing

Implementing an FM model

Improved factorization machines for predictive analytics

Neural factorization machines

Dataset description

Using NFM for movie recommendations

Model training

Model evaluation

Summary

11. Using Reinforcement Learning for Predictive Analytics

Reinforcement learning

Reinforcement learning in predictive analytics

Notation, policy, and utility in RL

Policy

Utility

Developing a multiarmed bandit's predictive model

Developing a stock price predictive model

Summary

Index