 |
View c6ea2926
- randomSelection function added. - crissCrossValidate and crissCrossPlot contributions by Harry Robertson added and harmonised to ClassifyR code style. - selectionMethod and classifier defaults become "auto" for crossValidate. Previously only documented but not implemented in code. - predict method for standalone use now finds the correct prediction method for each trained model.Dario Strbenac authored on 08/12/2022 06:15:07 |
|
View 989c4fce
- finalModel accessor added for getting final model properly. - Conversion back into selected features' original names fixed if feature selection does subsetting after the features are selected.Dario Strbenac authored on 15/09/2022 04:30:03 |
|
View b226d9f2
- Provide a convenience S4 method for ClassifyResult input object for ROCplot, so that users don't need to put their result into a list, like performancePlot, rankingPlot, selectionPlot already provide.Dario Strbenac authored on 13/09/2022 02:00:04 |
|
View 127675d5
- getFeatures functions added to simple params settings to extract selected features from within trained model where relevant. - Nearest Shrunken Centroid added as a simple params function and a classifier keyword option.Dario Strbenac authored on 25/08/2022 05:15:03 |
|
View 253c56fc
- All references to runTest and runTests in examples and vignette converted to crossValidate. End users should always use crossValidate from now on. - Minor fixes to code mistakes. - Performance tuning of training method parameters chosen within feature selection is now faithfully used in the model training.Dario Strbenac authored on 17/08/2022 13:55:15 |
|
View d37fb9c5
- Classifiers and feature selection functions no longer have multiple signaures and are private. - prepareData function to filter and subset input data using common ways, such as missingness and variability. - The variable renaming and storage in Original Feature and Renamed Feature reverted back to column metadata and assay / feature colums. - sampleInfo now reverted back to clinical.Dario Strbenac authored on 14/08/2022 23:45:28 |
|
View c6a73fc6
available is a new function to list the selection, classifier and multiView methods available for the user to choose from.Dario Strbenac authored on 08/08/2022 05:30:31 |
|
View f96a2df8
Default values of characteristicsList[['x']] and performanceName changed to "auto" so that characteristic which varies between results the most is automatically chosen. C-index is chosen for performance metric if results have survival predictions and balanced accuracy if classification was done.Dario Strbenac authored on 04/08/2022 01:35:47 |
|
View 00cff98e
- Data sets and feature names are converted into safe names such as Dataset1 and Feature1 for use inside of feature selection and classification functions. ClasifyResult class gains a featuresInfo slot which stores original identifiers as well as sanitised ones. - elasticNetPreval defunct parameter set removed.Dario Strbenac authored on 12/07/2022 14:35:03 |
|
View 4a3dc827
Double colons used for DataFrame function so classification works when parallelised on Windows. - Missing usage section restored by adding @rdname to all of the non-class S4 methods.Dario Strbenac authored on 11/03/2022 07:08:44 |
|
View fae56ddf
Repair of Roxygen documentation completed. No warnings or errors (from functions I wrote).Dario Strbenac authored on 10/03/2022 11:45:30 |
 |
View a770cc3e
Transitioned manuals to roxygen.Ellis Patrick authored on 14/01/2022 10:15:31 |
|
View ded9bca1
ROCplot can now either merge all predictions into one prediction set or consider each set separately, average them, and calculate confidence bands.Dario Strbenac authored on 10/12/2021 13:30:02 |
|
View a0cbbc14
- Major changes to driver functions and classes. Two new classes introduced CrossValParams and ModellingParams to store together parameters about cross-validation and data modelling. - Feature selection functions converted to feature ranking functions. Simplifies their parameters because they no longer need to know about cross-validation. Parameter tuning, such as the choice of top-p features, is done based on the rankings they produce. - Varieties are gone. Previously, a classifier could make multiple kinds of predictions for a single set of parameters, but it got too complicated to maintain.Dario Strbenac authored on 01/12/2021 05:05:48 |
|
View 9fa7bbd2
- Major changes to function definitions. None of them require fixed annotations like datasetName and classificationName any longer but ClassifyResult stores a table of characteristics which can be any pairs of characteristic and value. Plotting functions can group by any characteristic now. - performances no longer require the user to specify higher or lower is better. That detail is stored internally in a lookup table.Dario Strbenac authored on 05/11/2021 19:34:35 |