yejinjkim/supervise-phenotyping

MATLABBSD-2-Clause

Discriminative and Distinct Phenotyping by Nonnegative Tensor Factorization

DEC06 2016 Ver0.0

This is a repository of code for Discriminative and Distinct Phenotyping by Nonnegative Tensor Factorization.

REQUIREMENTS

This code requires the following:

Matlab (www.mathworks.com)
Tensor Toolbox Version 2.6 (http://www.sandia.gov/~tgkolda/TensorToolbox/index-2.6.html)

DATA FORMAT

count.csv: co-occurrence count of diagnosis and prescription for each subject.

  count, subject_id, diagnosis_id, prescription_id
  1, 35, 16, 41
  2, 35, 16, 68
  ...

label.csv : subject id of clinical outcome (eg. mortality, readmission) is 1 (eg. deceased, readmitted)
```
  subject_id
  2
  5
  10
  45
  ...
```

similarities.csv: pairwise similarity matrix of diagnoses and prescriptions.

  					d_diabetes, d_hypertension, ..., p_propofol, p_morphine_sulfate, ...
  d_diabetes				1,				0.3,				
  d_hypertension			0.3, 			1,					0.2
  ...
  p_propofol
  p_morphine_sulfate
  ...

RUN

Prepare CSV files: count.csv, label.csv, similarities.csv
Run main.m