4 projects
mechanism-learn
Mechanism-learn is a simple method to deconfound observational data such that any appropriate machine learning model is forced to learn predictive relationships between effects and their causes, despite the potential presence of multiple unknown and unmeasured confounding. The library is compatible with most existing ML deployments. The library is compatible with most existing ML deployments such as models built with Scikit-learn and Keras.
causal-sampler
causal-sampler is a python package that integrates multiple causal sampling techniques, e.g., causal bootstrapping and causally weighted Gaussian Mixture Models, offering standardized pipeline and interfaces.
causalbootstrapping
CausalBootstrapping is an easy-access implementation and extention of causal bootstrapping (CB) technique for causal analysis. With certain input of observational data, causal graph and variable distributions, CB resamples the data by adjusting the variable distributions which follow intended causal effects.
PFFRA
An Interpretable Machine Learning technique to analyse the contribution of features in the frequency domain. This method is inspired by permutation feature importance analysis but aims to quantify and analyse the time-series predictive model's mechanism from a global perspective.
