ashlinharris / mldemo Goto Github PK

• Condition names are used directly as symbols, which fails for names containing spaces. Some are also too long to include on figures.
• Direct usage implies tacit approval of terms.

Condition codes should be used in analysis, rather than descriptions. Conditions might correspond with multiple codes, so this should be addressed.

What's the easiest way to convert an OMOP format file to a Julia DataFrame?

The pipeline should handle zeroes and missing entries gracefully. Ideally, helpful warnings should be displayed. At the very least, the pipeline should not crash on these edge cases.

Certain functions still have hard-coded values by mistake. Many other functions can be generalized in an obvious way.

Most libraries should be included as a part of PreprocessMD.jl.

Currently, config.ini contains the directory names from my own machine, with Windows directory syntax. Is there a simple way to upload a dummy config file to the repository while keeping my own local config file in its place?

Currently, allergies are used as ML features, primarily due to ease of use. Conditions could be used instead.

Ideally, the meaning of any figure should be obvious to an unfamiliar audience. Figures should be appropriately labelled, perhaps with generated captions.

Systems that store medical data are often not connected to the internet. Can package dependencies be reduced or eliminated? Can the pipeline be embedded in a container?