strafe | novel architecture for modeling time

strafe is a Python library. strafe has no bugs, it has no vulnerabilities and it has low support. However strafe build file is not available and it has a Non-SPDX License. You can download it from GitHub.

STRAFE is a novel architecture for modeling time-series clinical data and predicting time-to-event. It is designed to provide better accuracy and interpretability than other models for this type of data, and can train on censored data. The STRAFE algorithm was applied to real-world claims data in the OMOP common data model (CDM) format. The STRAFE architecture is an expansion of SARD, a Transformer-based architecture developed by Kodiolam et al. [1], which takes as input time-series OMOP CDM data built from claims data to predict clinical outcomes. The expansion of the SARD model to time-to-event prediction was inspired by Hu et al. [4]. This repository includes the implementation of the STRAFE algorithm from the pre-process phase until the prediction phase. We also included the implementation of SARD and logistic regression for risk prediction and the survival baselines: random survival forest (RSF) [2] and DeepHit [3].