Skip to content

Latest commit

 

History

History
76 lines (56 loc) · 3.24 KB

NEWS.md

File metadata and controls

76 lines (56 loc) · 3.24 KB
Raw

Changelog

mldr 0.4

  • Reimplementation of performance evaluation metrics
  • Customizable treatment of undefined values in evaluation metrics
  • Support for single-quoted/double-quoted attributes in ARFF files
  • Support for stringsAsFactors
  • Support for negative -C parameter in MEKA header
  • API changes: The return value of mldr_evaluate has the same structure but different member names.

mldr 0.3.22

  • Improve how we load available MLD's in mldr()
  • Update CITATION file

mldr 0.3.18

  • Fix bug #21 when reading sparse datasets introducing zeroes instead of NAs
  • Fix #23 by ignoring case in @attribute tags
  • Fix bug #24 in filtering function
  • Improves parsing of attributes to correctly manage escaped apostrophes
  • Fix calculation of SCUMBLE CV to prevent NaN values
  • Optimizes some calculations (mean and CV of SCUMBLE are now 4x faster)
  • Export read.arff function to allow reading multilabel data without calculating measures
  • Add ability to load datasets from the mldr.datasets package in mldr() function

mldr 0.2.82

  • Fix bug in demo code for rebuilding an mldr object
  • Add global scumble.cv measure and SCUMBLE.CV measure per label
  • Add new concurrence module to ease analysis of concurrence among imbalanced labels
  • Display an analysis of concurrence between labels within the GUI
  • Add remedial preprocessing algorithm
  • Fix bug #18 when reading ARFF with capitalized '@relation' parameters
  • A Citation file (accesible from citation("mldr")) has been added

mldr 0.2.51

  • Fix bug #14 when reading certain sparse datasets (e.g. Yahoo)
  • Support for multiple plot types in one call
  • Add num.inputs measure for input attributes
  • Add color parameters for plotting functions

mldr 0.2.33

  • New redesigned GUI to ease usability
  • Add ability to read MEKA datasets from GUI

mldr 0.2.25

New features

  • New mldr_evaluate() function to assess multilabel classifier performance. Taking as input an mldr object and a matrix with predictions this function returns a list of metrics, including Accuracy, AUC, AveragePrecision, Coverage, FMeasure, HammingLoss, MacroAUC, MacroFMeasure, MacroPrecision, MacroRecall, MicroAUC, MicroFMeasure, MicroPrecision, MicroRecall, OneError, Precision, RankingLoss, Recall, and SubsetAccuracy
  • Added more parameters to mldr function so labels can be identified via specific names, indices or their count.
  • Added vignette

Fixes

  • Fixed call to chordDiagram for newer versions of the circlize package.
  • Fixed imports to avoid NOTEs on devel builds.
  • Fixed parameters in calls to pROC functions.

mldr 0.1.70

First release of mldr. This version includes:

  • Ability to read multi-label data sets from ARFF and XML files in Mulan or MEKA format.
  • Ability to write to ARFF and XML files in both Mulan and MEKA formats.
  • Different ways to display data and relevant measures from data sets.
  • Several plots for multi-label data visualisation.
  • Functions to operate with mldr objects: filtering, joining and structure comparison.
  • BR and LP transformations of multi-label datasets.
  • Ability to create new mldr objects out of data.frames.
  • Sample datasets: birds, emotions and genbase.