English | MP4 | AVC 1280×720 | AAC 48KHz 2ch | 1h 10m | 203 MB
Ensembles involve groups of models working together to make more accurate predictions. When creating complete deployed solutions, data scientists may also leverage passing data from one model to another or using models in combination—also known as metamodeling. These techniques are dominant among winners of modeling competitions like Kaggle as well as leading data science teams around the world. In this advanced course, you can learn how to add ensembles and metamodeling to your toolset. Instructor Keith McCormick provides a conceptual introduction that can be applied in any program: R, Python, SPSS, or SAS. He introduces the most essential ensemble algorithms and explains the basics of metamodeling. Plus, review two case studies that show how to combine supervised and unsupervised ensembles and how to route subpopulations of data to different models in a metamodeling scenario.
Topics include:
- What is an ensemble?
- Types of ensembles
- Measuring model accuracy
- Boosting, bagging, and stacking
- Visualizing bias and variance
- Important and influential ensemble algorithms
- Metamodeling
Table of Contents
1 The most accurate machine learning models
2 What you should know
3 Ensemble wins Netflix Prize
4 What is an ensemble
5 Types of models and modeling algorithms
6 Types of ensembles
7 Measuring model accuracy Value estimation
8 Understanding model error Classification
9 Stacking
10 Voting for classification
11 Error decomposition
12 Visualizing bias and variance
13 Curse of dimensionality
14 Is Occam’s Razor always true
15 What is Bootstrap aggregating
16 What is Boosting and how does it work
17 Gradient boosting demo
18 Random forest
19 Model search by bumping
20 AdaBoost, XGBoost, Light GBM, CatBoost
21 Super Learner, Subsemble, StackNet
22 What are people working on now
23 Combining supervised and unsupervised
24 Routing cases to different models
25 Next steps
Resolve the captcha to access the links!