Automated machine learning¶
DSS contains a powerful automated machine learning engine that allows you to get highly optimized models with minimal intervention.
At your choice, in DSS, you can select between:
- Having the full control over all training settings, algorithm settings and optimization process, including writing your own custom models and using advanced deep learning models
- Using DSS powerful automatic machine learning engine in order to effortlessly get models
In addition to algorithms selection and optimization, the automated machine learning performs:
- Automatic features handling, including handling of categorical and text variables, handling of missing values, scaling, …
- Semi-automatic massive features generation
- Optional features selection
- Go to the Flow for your project
- Click on the dataset you want to use
- Select the Lab
- Select Quick model then Prediction
- Choose your target variable (one of the columns) and Automated Machine Learning
- Select one of the prediction styles
The Automated Machine Learning engine allows you to choose between three main prediction styles
When selecting this prediction style, DSS will select a variety of models, privileging variety and speed over pure performance. The main goal of this is to quickly give you first results. It will help you decide whether you need to go further with more advanced models, or if you should first do some more feature engineering
This prediction style is focused on giving “white-box” models for which it is easier to understand the predictions and the driving factors.
DSS will choose both decision trees and linear models.
Training is generally quick.
When selecting this prediction style, DSS will select a variety tree-based models with a very deep hyper-parameter optimization search. This will in most cases give the best possible prediction performance, at the expanse of interpretability.
Training time will be strongly increased when choosing this prediction style
DSS can compute interactions between variables, such as linear and polynomial combinations. These generated features allow for linear methods, such as linear regression, to detect non-linear relationship between the variables and the target. These generated features may improve model performance in these cases.
See Prediction settings for more information