Low-Code AI: A Practical Project-Driven Introduction to Machine Learning

Preface
Who Should Read This Book?
What Is and Isn’t in This Book
Conventions Used in This Book
Using Code Examples
O’Reilly Online Learning
How to Contact Us
Acknowledgments
1. How Data Drives Decision Making in Machine Learning
What Is the Goal or Use Case?
An Enterprise ML Workflow
Defining the Business Objective or Problem Statement
Data Collection
Data Preprocessing
Data Analysis
Data Transformation and Feature Selection
Researching the Model Selection or Using AutoML (a No-Code Solution)
Model Training, Evaluation, and Tuning
Model Testing
Model Deployment (Serving)
Maintaining Models
Summary
2. Data Is the First Step
Overview of Use Cases and Datasets Used in the Book
1. Retail: Product Pricing
2. Healthcare: Heart Disease Campaign
3. Energy: Utility Campaign
4. Insurance: Advertising Media Channel Sales Prediction
5. Financial: Fraud Detection
6. Energy: Power Production Prediction
7. Telecommunications: Customer Churn Prediction
8. Automotive: Improve Custom Model Performance
Data and File Types
Quantitative and Qualitative Data
Structured, Unstructured, and Semistructured Data
Data File Types
How Data Is Processed
An Overview of GitHub and Google’s Colab
Use GitHub to Create a Data Repository for Your Projects
1. Sign up for a new GitHub account
2. Set up your project’s GitHub repo
Using Google’s Colaboratory for Low-Code AI Projects
1. Create a Colaboratory Python Jupyter Notebook
2. Import libraries and dataset using Pandas
3. Data validation
4. A little bit of exploratory data analysis
Summary
3. Machine Learning Libraries and Frameworks
No-Code AutoML
How AutoML Works
Machine Learning as a Service
Low-Code ML Frameworks
SQL ML Frameworks
Google’s BigQuery ML
Amazon Aurora ML and Redshift ML
Open Source ML Libraries
AutoKeras
Auto-Sklearn
Auto-PyTorch
Summary
4. Use AutoML to Predict Advertising Media Channel Sales
The Business Use Case: Media Channel Sales Prediction
Project Workflow
Project Dataset
Exploring the Dataset Using Pandas, Matplotlib, and Seaborn
Load Data into a Pandas DataFrame in a Google Colab Notebook
Explore the Advertising Dataset
Descriptive analysis: Check the data
Explore the data
Heat maps (correlations)
Scatterplots
Histogram distribution plot
Export the advertising dataset
Use AutoML to Train a Linear Regression Model
No-Code Using Vertex AI
Create a Managed Dataset in Vertex AI
Select the Model Objective
Build the Training Model
Evaluate Model Performance
Model Feature Importance (Attribution)
Get Predictions from Your Model
Summary
5. Using AutoML to Detect Fraudulent Transactions
The Business Use Case: Fraud Detection for Financial Transactions
Project Workflow
Project Dataset
Exploring the Dataset Using Pandas, Matplotlib, and Seaborn
Loading Data into a Pandas DataFrame in a Google Colab Notebook
Exploring the Dataset
Descriptive analysis
Exploratory analysis
Exporting the Dataset
Classification Models and Metrics
Using AutoML to Train a Classification Model
Creating a Managed Dataset and Selecting the Model Objective
Exploring Dataset Statistics
Training the Model
Evaluating Model Performance
Model Feature Importances
Getting Predictions from Your Model
Summary
6. Using BigQuery ML to Train a Linear Regression Model
The Business Use Case: Power Plant Production
Cleaning the Dataset Using SQL in BigQuery
Loading a Dataset into BigQuery
Exploring Data in BigQuery Using SQL
Using the Null function to check for null values
Using the Min and Max functions to determine acceptable data ranges
Saving query results using a DDL statement in BigQuery
Linear Regression Models
Feature Selection and Correlation
Google Colaboratory
Plotting Feature Relationships to the Label
The CREATE MODEL Statement in BigQuery ML
Using the CREATE MODEL statement
View evaluation metrics of the trained model
Using the ML.PREDICT function to serve predictions
Introducing Explainable AI
Explainable AI in BigQuery ML
Modifying the CREATE MODEL statement
Using the ML.GLOBAL_EXPLAIN function
Using the ML.EXPLAIN_PREDICT function to compute local explanations
Exercises
Neural Networks in BigQuery ML
Brief Overview of Neural Networks
Activation Functions and Nonlinearity
Training a Deep Neural Network in BigQuery ML
Exercises
Deep Dive: Using Cloud Shell to View Your Cloud Storage File
Summary
7. Training Custom ML Models in Python
The Business Use Case: Customer Churn Prediction
Choosing Among No-Code, Low-Code, or Custom Code ML Solutions
Exploring the Dataset Using Pandas, Matplotlib, and Seaborn
Loading Data into a Pandas DataFrame in a Google Colab Notebook
Understanding and Cleaning the Customer Churn Dataset
Checking and converting data types
Exploring summary statistics
Exploring combinations of categorical columns
Exploring interactions between numeric and categorical columns
Transforming Features Using Pandas and Scikit-Learn
Feature selection
Encoding categorical features using scikit-learn
Generalization and data splitting
Building a Logistic Regression Model Using Scikit-Learn
Logistic Regression
Training and Evaluating a Model in Scikit-Learn
Classification Evaluation Metrics
Serving Predictions with a Trained Model in Scikit-Learn
Pipelines in Scikit-Learn: An Introduction
Building a Neural Network Using Keras
Introduction to Keras
Training a Neural Network Classifier Using Keras
Building Custom ML Models on Vertex AI
Summary
8. Improving Custom Model Performance
The Business Use Case: Used Car Auction Prices
Model Improvement in Scikit-Learn
Loading the Notebook with the Preexisting Model
Loading the Datasets and the Training-Validation-Test Data Split
Exploring the Scikit-Learn Linear Regression Model
Feature Engineering and Improving the Preprocessing Pipeline
Looking for easy improvements
Feature crosses
Hyperparameter Tuning
Hyperparameter tuning strategies
Hyperparameter tuning in scikit-learn
Model Improvement in Keras
Introduction to Preprocessing Layers in Keras
Creating the Dataset and Preprocessing Layers for Your Model
Building a Neural Network Model
Hyperparameter Tuning in Keras
Hyperparameter Tuning in BigQuery ML
Loading and Transforming Car Auction Data
Training a Linear Regression Model and Using the TRANSFORM Clause
Configure a Hyperparameter Tuning Job in BigQuery ML
Regularization
Using hyperparameter tuning in the CREATE MODEL statement
Options for Hyperparameter Tuning Large Models
Vertex AI Training and Tuning
Automatic Model Tuning with Amazon SageMaker
Azure Machine Learning
Summary
9. Next Steps in Your AI Journey
Going Deeper into Data Science
Working with Unstructured Data
Working with image data
Working with text data
Generative AI
Explainable AI
ML Operations
Continuous Training and Evaluation
Summary
Index