The code in this repo converts logistic regression models originally created in R to study Ebola diagnosis and prognosis into TensorFlow models, and save them in TensorFlow Lite format to load into ClinicApp.
RCS terms are included into the TensorFlow model using a custom Keras layer where the RCS formula is implemented, and then incorporated as an input layer. See the ebola-pediatric-prognosis notebook for details.
This repo also includes code demonstrating how to calculate and visualize the SHAP values of a model for a given data instance, see the shap-values-heart-disease notebook.