Being tired of defining the same lines of code every time that I use specific things, I decided to implement these in this package, so that I just need to import the particular tools that I need.
Up to now, the tools implemented are:
- A module for principal component regression (and classification). This module includes an object that simply builds either a Pipeline or a GridSearchCV with a pipeline that concatenates a Variance Threshold feature selection, Scaling, PCA and a linear regressor with regularisation (None, L1, L2 or ElasticNet). Alternatively, there are also coded LassoPCR, LogisticPCR, RidgePCR and ElasticnetPCRobjects that in principle should be much faster than using GridSearchCV. More kinds of these classes might be added in the future.
- A stratifiedKFold and RepeatedStratifiedKFold objects for regression.
- A module for connectivity-based predictive modelling.
- A module for transmodal learning.
- A Columwise transformer in the module preprocessing that replaces each column with the cross-validated predictions using a Linear Regression model.
Under continuously development.
git clone https://github.com/jrasero/my-scikit-tools.git
cd my-scikit-tools
pip install -U .