The imputation and model pipelines for the INFORM Severity data.
Ridge Regression with 88% accuracy(alpha=25 and 10-fold CV)
LightGBM with 99% accuracy (10 fold CV and reg_alpha: 1.96248962962756, reg_lambda: 4.688549449064171)
There is a 1 single pipeline for FeatureEngineering, Categorical encoding, Transformations and Imputations.
We just need to deploy the final_pipelines for both the models stored in file: final_pipe.pkl
Inference API (serverless) does not yet support transformers models for this pipeline type.