README.md

In hestiri/mlho: Machine Learning for Health Outcomes (MLHO)

Hello MLHO!

MLHO (pronounced as melo) is a thinkin’ Machine Learning framework that implements iterative sequential representation mining, and feature and model selection to predict health outcomes.

Installation

You can install the released version of mlho from Github with:

devtools::install_github("hestiri/mlho")

Data model

To implement MLHO you’ll need 2 tables, which can be extracted from any clinical CMD. The current examples are based on the i2b2 star schema.

1- a table with outcome labels (called labeldt) and patient numbers

| | | | :----------- | :----- | | patient_num | label | | character | factor |

2- a patient clinical data table (called dbmart) with 3 columns. Concepts are used as features by MLHO.

| | | | | :----------- | :---------- | :-------- | | patient_num | start_date | phenx | | character | date | character |

The column phenx contains the entire feature space. In an i2b2 data model, for instance, this column is the equivalent of concept_cd.

3- a demographic table is optional, but recommended.

| | | | | | :----------- | :-------- | :-------- | :-------- | | patient_num | age | gender | … | | character | character | character | character |

see the instructions on how to use the MLHO package on the articles page

hestiri/mlho documentation built on March 20, 2023, 11:04 p.m.