language: en
tags:
- fill-mask
license: mit
Datasets used to test hospitalization prediction models
Datasets used in testing the trained models presented in the paper "Predicting Hospitalization with LLMs from Health Insurance Data". We provide data in English and Portuguese.
All data here is anonymized and represents the patient's event history, ordered from left to right in a sentence format, with data on the left being oldest and data on the right being newest. Label values 0 indicate that the patient was not hospitalized after that historical sequence, and 1 otherwise.
The original data is in Portuguese and was translated into English using the translation-pt-en-t5 translation model. The data were randomly selected 5 times using the seeds [12, 23, 34, 45, 56] for each selection. Therefore, for each of the versions, Portuguese and English, there are 5 test sets.
Dataset with the suffix "pt" is in Portuguese and with the suffix "en" in English.
Test models
To test the models, use the guidelines as described in the paper "Predicting Hospitalization with LLMs from Health Insurance Data". To carry out tests with RandomForest it is necessary to extract the embeddings using the sentence transformers framework.
More Information
Refer to the original paper, Predicting Hospitalization with LLMs from Health Insurance Data
Refert to another article related to this research, Predicting Hospitalization from Health Insurance Data
Questions?
Email:
- Everton F. Baro: efbaro@inf.ufpr.br, everton.barros@ifpr.edu.br
- Luiz S. Oliveira: luiz.oliveira@ufpr.br
- Alceu de Souza Britto Junior: alceu.junior@pucpr.br