Data Mining Methods

We define the step to go from the features to a radiomics model, whether it is classification, regression, or survival, as data mining. Within this step, combined model selection and hyperparameter optimization is performed to find the best performing workflow, i.e. combination of models and hyperparameters.

In this part of the documentation, all methods available in the data mining step are described in the order they occur. The actual single workflow fitting and scoring is done in WORC.WORC.classification.fitandscore.fit_and_score, in which all of these methods are embedded.

Here, we provide a rationale for the methods. For a comprehensive overview of all functions and parameters, please look at the config chapter.

OneHotEncoding

Documentation WIP.

Imputation

Documentation WIP.

Feature Selection: Groupwise

Documentation WIP.

Feature Scaling

Documentation WIP.

Feature Selection: Variance

Documentation WIP.

Feature Selection: Univariate Statistical Test

Documentation WIP.

Feature Selection: Relief

Documentation WIP.

Feature Selection: Select from model

Documentation WIP.

Dimensionality Reduction: principal component analysis (PCA)

Documentation WIP.

Resampling

Documentation WIP.

Machine Learning

Documentation WIP.

Classification

Documentation WIP.

Regression

Documentation WIP.

Survival

Documentation WIP.