Tabular¶
For standard datasets that are represented as tables (stored as CSV file, parquet from database, etc.), AutoGluon can produce models to predict the values in one column based on the values in the other columns. With just a single call to fit(), you can achieve high accuracy in standard supervised learning tasks (both classification and regression), without dealing with cumbersome issues like data cleaning, feature engineering, hyperparameter optimization, model selection, etc.
5 min tutorial on fitting models with tabular datasets.
Essential information about the most important settings for tabular prediction.
In-depth tutorial on controlling various aspects of model fitting.
Modeling data tables with image, text, numeric, and categorical features.
AutoGluon’s default feature engineering and how to extend it.
How to predict multiple columns in a data table.
Using AutoGluon for Kaggle competitions with tabular data.
How to train models with GPU support.
How to add a custom metric to AutoGluon.
How to add a custom model to AutoGluon.
How to add a custom model to AutoGluon (Advanced).
Tutorial on optimizing the predictor artifact for production deployment.