Modeling
To train machine learning models and perform probability predictions based on data, the ANTAVIRA platform implements a linear modeling module that includes the following machine learning algorithms:
- Logistic regression;
- Linear redression;
- Decision tree;
- Decision forest.
In the linear modeling module, in addition to choosing a machine learning algorithm, you are also invited to configure:
- Performing gradient descent if necessary, including choosing the type of descent (ascending; descending), selection rules for variables (Log Likelihood; Error 1,2 kind; Error 1 kind; Error 2 kind; AUC) and setting the number of iterations.
- Selection of variables by IV / IG (in developing) so that the platform selects for modeling only those variables whose significance for the objective function exceeds the value you specified.
- Trained value (in developing) (if at the stage of processing the values of variables you have chosen a method associated with the transformation of features into categorical ones: grouping, clustering, binning).
Logistic regression
When building a machine learning model using logistic regression, you are offered to set the following parameters:
- Regularization;
- Maximum number of iterations;
- Optimization algorithm;
- Penalty for error.
Linear regression
The parameters that enable conducting experiments are currently under development.
Decision tree
The parameters that enable conducting experiments are currently under development.
Decision forest
The parameters that enable conducting experiments are currently under development.
