Data manipulation

This section provides examples of common tasks performed when preparing data for machine learning. These examples are run on a local cluster.

Note

The examples in this section include datasets that are pulled from GitHub and S3.

• Upload a file
• Import a file
• Import multiple files
• Download data
• Change the Column Type
• Combine columns from two datasets
• Combine rows from two datasets
• Fill NA values
• Group by
• Impute data
• Merge two datasets
• Pivot tables
• Replace values in a frame
• Slice columns
• Slice rows
• Sort columns
• Split datasets into training/testing/validating
• Tokenize strings

Feature engineering

H2O-3 also has methods for feature engineering. Target Encoding is a categorical encoding technique which replaces a categorical value with the mean of the target variable (this is especially useful for high-cardinality features). Word2vec is a text processing method which converts a corpus of text into an output of word vectors.