Example use cases

Tabular data

Kaggle Heart Failure Prediction

This is a binary classification problem where the goal is to predict risk of heart disease. The heart failure dataset is available on Kaggle. The dataset contains 11 features and 1 target variable.

• Heart Failure Prediction
  • Import Libraries
  • Constants
  • Data Loading
    • Sex values
    • Age distribution
    • Outcome distribution
    • Identifying feature types
    • Creating data preprocessors
  • Creating Hugging Face Dataset
  • Model Creation
  • Task Creation
  • Training
  • Prediction
  • Evaluation
  • Report Generation

MIMICIV Mortality Prediction

This is a binary classification problem where the goal is to predict risk of in-hospital mortality. The MIMICIV dataset is an EHR dataset collected from a single hospital site, which includes ICU data.

• Mortality Prediction
  • Import Libraries
  • Constants
  • Data Querying & Processing
    • Compute mortality (labels)
  • Data Inspection and Preprocessing
    • Drop NaNs based on the NAN_THRESHOLD
    • Outcome distribution
    • Gender distribution
    • Age distribution
    • Identifying feature types
    • Creating data preprocessors
  • Creating Hugging Face Dataset
  • Model Creation
  • Task Creation
  • Training
  • Prediction
  • Evaluation
  • Report Generation

Synthea Prolonged Length of Stay Prediction

This is a binary classification problem where the goal is to predict whether a patient will have a prolonged length of stay in the hospital (more than 7 days). The synthea dataset is generated using Synthea which is a synthetic patient generator. The dataset contains observations, medications and procedures as features.

• Prolonged Length of Stay Prediction
  • Import Libraries
  • Constants
  • Data Querying
    • Compute length of stay (labels)
  • Data Inspection and Preprocessing
    • Drop NaNs based on the NAN_THRESHOLD
    • Length of stay distribution
    • Outcome distribution
    • Gender distribution
    • Age distribution
    • Identifying feature types
    • Creating data preprocessors
  • Creating Hugging Face Dataset
  • Model Creation
  • Task Creation
  • Training
  • Prediction
  • Evaluation
  • Report Generation

Diabetes 130-US Hospitals for Years 1999-2008 Readmission Prediction

This is a binary classification problem where the goal is to predict risk of readmission. The diabetes dataset is available on UCI Machine Learning Repository. The dataset contains 47 features and 1 target variable.

• Readmission Prediction
  • Install libraries
  • Import Libraries
  • Constants
  • Data Loading
    • Sex values
    • Age distribution
    • Outcome distribution
    • Identifying feature types
    • Creating data preprocessors
  • Creating Hugging Face Dataset
  • Model Creation
  • Task Creation
  • Training
  • Prediction
  • Evaluation
  • Report Generation

Image data

NIH Chest X-ray classification

This tutorial showcases the use of the tasks API to implement a chest X-ray classification task. The dataset used is the NIH Chest X-ray dataset, which contains 112,120 frontal-view X-ray images of 30,805 unique patients with 14 disease labels.

The tutorial also demonstrates the use of the evaluate API to evaluate the performance of a model on the task.

• Chest X-Ray Disease Classification
  • Import Libraries
    • Generate Historical Reports
    • Initialize Periodic Report
  • Load Dataset
    • Model Creation
  • Multilabel AUROC by Pathology and Sex
  • Multilabel AUROC by Pathology and Age
  • Log Performance Metrics as Tests w/ Thresholds
    • Populate Model Card Fields