Machine Learning Engineering (Zoomcamp) - Data Talks

Chapters

Chapter 1: Introduction to Machine Learning

• 1.1 - Introduction to Machine Learning

• 1.2 - Machine Learning vs Rule-Based Systems

• 1.3 - Supervised Machine Learning

• 1.4 - CRISP-DM

• 1.5 - Model Selection

• 1.6 - Environment

• 1.7, 1.8, 1.9 - Introduction to Numpy, Pandas and Linear Algebra Refresh

Chapter 2: Machine Learning for Regression

Project plan:

* Prepare data and Exploratory data analysis (EDA)
* Use linear regression for predicting price
* Understanding the internals of linear regression
* Evaluating the model with RMSE
* Feature engineering
* Regularization
* Using the model

• 2.1 - Car price prediction project

• 2.2 - Data preparation

• 2.3 - Exploratory Data Analysis (EDA)

• 2.4 - Setting Up The Validation Framework

• 2.5 - Linear regression

• 2.6 - Linear regression vector form

• 2.7 - Training linear regression: Normal equation

• 2.8 - Baseline model for car price prediction project

• 2.9 - Root mean squared error

• 2.10 - Using RMSE on validation data

• 2.11 - Feature engineering

• 2.12 - Categorical variables

• 2.13 - Regularization

• 2.14 - Tuning the model

• 2.15 - Using the model

Chapter 3: Machine Learning for Classification

• 3.1 - Churn Prediction Project

• 3.2 - Data Preparation

• 3.3 - Setting Up The Validation Framework

• 3.4 - EDA

• 3.5 - Feature Importance: Churn Rate And Risk Ratio

• 3.6 - Feature Importance: Mutual Information

• 3.7 - Feature Importance: Correlation

• 3.8 - One-Hot Encoding

• 3.9 - Logistic Regression

• 3.10 - Training Logistic Regression with Scikit-Learn

• 3.11 - Model Interpretation

• 3.12 - Using the Model

• 3.13 - Summary

  • Feature importance - risk, mutual information, correlation

  • One-hot encoding can be implemented with DictVectorizer

  • Logistic regression - linear model like linear regression

  • Output of log reg - probability

  • Interpretation of weights is similar to linear regression

Chapter 4: Evaluation Metrics for Classification

• 4.1 Evaluation metrics: session overview

• 4.2 Accuracy and dummy model

• 4.3 Confusion table

• 4.4 Precision and Recall

• 4.5 ROC Curves

• 4.6 ROC AUC

• 4.7 Cross-Validation

Chapter 5 - Deploying Machine Learning Models

• 5.1 Intro / Session overview

• 5.2 Saving and loading the model

• 5.3 Web services: introduction to Flask

• 5.4 Serving the churn model with Flask

• 5.5 Python virtual environment: Pipenv

• 5.6 Environment management: Docker

• 5.7 Deployment to the cloud: AWS Elastic Beanstalk (optional)

Chapter 6 - Decision Trees and Ensembles Learning

• 6.1 Credit risk scoring project

• 6.2 Data cleaning and preparation

• 6.3 Decision trees

• 6.4 Decision tree learning algorithm

• 6.5 Decision trees parameter tuning

• 6.6 Ensemble learning and random forest

• 6.7 Gradient boosting and XGBoost

• 6.8 XGBoost parameter tuning

• 6.9 Selecting the best model

Chapter 7 - Deploy with BentoML

• 7.1 Intro/Session Overview

• 7.2 Building Your Prediction Service with BentoML

• 7.3 Deploying Your Prediction Service

• 7.4 Sending, Receiving and Validating Data

• 7.5 High-Performance Serving

• 7.6 Bento Production Deployment

Chapter 8 - Deep Learning

• 8.1 Fashion classification

• 8.1b Setting up the Environment on Saturn Cloud

• 8.2 TensorFlow and Keras

• 8.3 Pre-trained convolutional neural networks

• 8.4 Convolutional neural networks

• 8.5 Transfer learning

• 8.6 Adjusting the learning rate

• 8.7 Checkpointing

• 8.8 Adding more layers

• 8.9 Regularization and dropout

• 8.10 Data augmentation

• 8.11 Training a larger model

• 8.12 Using the model

Chapter 9 - Serveless Deep Learning

• 9.1 Introduction to Serverless

• 9.2 AWS Lambda

• 9.3 TensorFlow Lite

• 9.4 Preparing the code for Lambda

• 9.5 Preparing a Docker image

• 9.6 Creating the lambda function

• 9.7 API Gateway: exposing the lambda function