Files from the Practical Machine Learning on H2O course on Coursera
- Week 1 - Deep Learning Example: Introduction to H2O using deep learning
- Week 1 - Auto ML Example: Introduction to H2O's Auto ML functionality
- Week 2 - Random Forest Example: Example using Iris and a random forest
- Week 2 - GBM Example: Example using Iris and a GBM
- Week 2 - Importing Data: Sample code importing artificial data
- Week 2 - Overfitting: Examples using train/valid/test and cross validation showing the effects of overfitting a model
- Week 2 - Assignment: Assignment to create artificial data and create a default model and an overfitting model
- Week 3 - GLMs for Exploration: Example using GLM to do data exploration on smoking data
- Week 3 - Naive Bayes: Example using Naive Bayes on Iris
- Week 3 - Data Manipulation: Examples of how to manipulate data within H2O using airlines data
- Week 3 - Grid Search: Examples of performing grid searches on a GLM using airlines data
- Week 4 - Binding and Merging Data: Examples of how to bind rows and columns and merge frames
- Week 4 - Deep learning: Examples of how to build deep learning models and adjust parameters using airlines data
- Week 4 - Grid Search with Deep Learning: Examples of how to use a grid search with deep learning using airlines data
- Week 4 - Deep Learning Regression: Examples of how to use deep learning to also build regression models using airlines data
- Week 4 - Assignment: Assignment to create a classification model using deep learning, comparing a default model to a tuned model using cacao dataset
- Week 5 - Handling Missing Data: Example strategies for dealing with missing data using the airlines data
- Week 5 - PCA and GLRM: Examples of how to use PCA and GLRM with the iris dataset
- Week 5 - Autoencoders: Examples of using autoencoders, deep features, and anomalies on the iris dataset
- Week 6 - Stacked Ensembles: Examples building stacked ensembles using the airlines data
- Week 6 - Assignment: Assignment to create an ensemble model using four models of at least three different model types on a house price dataset