mikropml: User-Friendly R Package for Supervised Machine Learning Pipelines

Published May 13, 2021 | Version v1.0.0

Software Open

mikropml now has a logo created by @NLesniak!
Made documentation improvements (#238, #231 @kelly-sovacool; #256 @BTopcuoglu).
New option in preprocess_data(): prefilter_threshold (#240, @kelly-sovacool, @courtneyarmour).
- Remove any features that appear in N=prefilter_threshold or fewer rows in the data.
- Created function remove_singleton_columns() called by preprocess_data() to carry this out.
New option in get_feature_importance(): groups (#246, @kelly-sovacool).
- Provide custom groups of features to permute together during permutation importance.
- groups is NULL by default; in this case, correlated features above corr_thresh are grouped together.
preprocess_data() now replaces spaces in the outcome column with underscores (#247, @kelly-sovacool, @JonnyTran).
Clarify in the intro vignette that we do not support multi-label outcomes. (#254, @zenalapp)
Optional progress bar for preprocess_data() and get_feature_importance() using the progressr package (#257, @kelly-sovacool, @JonnyTran, @FedericoComoglio).
The mikropml paper is soon to be published in JOSS!

Files

Name	Size	Download all
SchlossLab/mikropml-v1.0.0.zip md5:68e581dd22cc96b1dcf36cdbb0b6b860	4.4 MB	Preview Download