- Jupyter Notebook 99.9%
| Filename | Latest commit message | Latest commit date |
|---|---|---|
| DNN | ||
| LSTM | ||
| ML | ||
| prediction_plot | ||
| valid_data_predict | ||
| README.md | ||
Prediction of adverse outcomes for poisoning patients
Apply interpretable artificial intelligence methods to predict the adverse outcomes of acute poisoning.
Overview
This repository contains Jupyter notebooks that implement an explainable artificial intelligence (AI) framework for predicting adverse outcomes—specifically, incomplete recovery or death—in patients with acute poisoning, while performing temporal external validation and providing transparent, clinician-oriented explanations.
Code list
a. DNN, LSTM
Data preprocessing and cleaning; Development and training of DNN and LSTM models; Prediction and outcome evaluation
b. ML
Data preprocessing and cleaning; Development and training of ML models; Prediction and outcome evaluation
c. prediction_plot
Evaluates the final model's predictive performance; Interprets predictions via SHAP analysis; Structured SHAP outputs for an LLM to generate patient-specific risk assessments; Clinical explanations, enhancing interpretability
d. valid_data_predict
Temporal External Validation of the Model