Togaware: Rattle: Change Log

freedom is in everyone's language

Frihed Vrijheid Liberté Freiheit Ελευθερία Свобода Bebas Libertad

Home
Services
Freedoms
Resources

- LaTeX

Supporting

- Analytics/IAPA

- AusDM

- PAKDD

Hosting

- Dirt Navigator

- Gallery

About Us

Rattle: Change Log

This is a brief Rattle change log. Each revision corresponds to a Rattle version. Generally, only major changes on each revision check-in are recorded here. The current Rattle version is 2.3.27. The version number is determined as 2.3.n where n is the current revision identified below minus 230.

257 080516 Fix up branding: textview updates; plot title; plot subtitles
           Only show web page link on "I" button in menu bar for Rattle.
           Autoload any supplied csv file.
256 080516 For regression/linear, use lm.
255 080515 Textviews and plots now identify the appname.
           Add the regression button again for Classification
           Ensure the glm family selector is shown.
           Start fixing help and messages to switch on appname.
           Fix references to the Variables tab (now the Select tab)
254 080513 Note that csvname as URL on startup not yet supported.
           Fix bug where hclust buttons had disappeared!
           Use URLdecode to ensure special characters are restored.
           Note Score and PrVOb not support more than 2 classes.
253 080511 Move to generalising the Rattle badge, accommodating RStat
           Remove redundant Info dialogs where status contains the same message
           Add new RStat option (rebadging plus eventually tuned functionality)
           Avoid scientific notation in printing numbers (using scipen)
           Support specifying a CSV file using a URL
           Identify LM as regression in the log, not logistic regression
           For pmml do not claim copyright for generated PMML
252 080430 Remove Lift filter on Show Rules in Assoc - arules bug?
           Fix bug when no target is selected (Peter Christen).
           Improve heuristic for selecting target (must have < 5 levels).
251 080428 Fix bug where paradigm change when Model/Eval exposed resizes window
           Fix bug where after building regression models extra Evals re-added
250 080427 Implement nnet for classification and regression.
           Ensure scores for nnet regression are working
           Export nnet to PMML
           Make unimplemented paradigms invisible, not greyed out.
           Hide/show score options on radio button toggle
           Hide elements not in use in general (HIG Guidelines)
           On change paradigm clear all old models
           Remove Boost, Forest, SVM from Regression
           Remove Linear from Classification (logistic == mulitnom, i.e. nnet)
249 080425 Score file can optionally contain all columns or just idents.
           Tidy up some error messages with the errorReport function.
248 080425 All scores are now placed into a single file.
           A filename to save to is now requested on saving scores.
247 080422 Change Normalise to Rescale.
           Add Wei Lin's rescale over a matrix operation.
           Update pmml.lm man to lessen the advertising. 
246 080421 Add pmml.arules to pmml and add Rattle support (Michael Hahsler)
           Fix pmml to export S3 methods, not full names. Update Rattle.
           Release version 1.1.7 of pmml.
245 080419 Spelling fix (Peter Christen)
           Rule generation by arules now works so use it (Peter Christen)
           Support .Rattle files.
244 080418 Bug Fix: use na.omit for prcomp (Esther)
           Bug Fix: zzz.R should not be using getenv any more.
243 080418 Use env RATTLE_SCORE for the name of score file to save to.
           Support Close and Quit (Close does not kill the R Console).
           Bug Fix: glm score when na.omit has effect => error.
           Remove getenv option. Just use the env var if it is defined.
242 080417 Add getenv option to rattle function for RATTLE_DATA.
           Change zzz .onLoad to .onAttach to auto start rattle if
           rattle package is autoladed i.e. getOption("defaultPackages")
           Cleanup pmml documentation code and add other authors to pmml.
           Fix up the generation of the package repository.
241 080415 On paradigm shift, go to Select or Data tab.
           Add initial support for Linear model regression vs classification
240 080414 Use supprt@togaware.com
           Move severl var@attr to attr(var, "attr") for non-S4 objects
           Change focus to Select tab on a Paradigm change
239 080413 Enable Regression and start setting up for this paradigm
           Currently working for rpart - build and score work.
238 080412 Bug fix: Error reading numeric matrix - bad use of & (Mark Andersen)
           Include anova output in glm display
           pmml.rpart check for class methd, else fail
           Add kernlab to pmml DESCRIPTION (Wen-Chi Lin)
237 080406 Bug fix: ByGroup of a singleton rank was set to 0 - now 99.
           Bug fix: BQ and BK binning generate factor, not ordered factor
                    since RF can't handle ordered factor.
           Bug fix: missing _ caused remap vars not identified as Input
236 080329 Bug fix: normalise by group when group has one element -> 0.
235 080328 Bug fix: normalise fix ordered factor problem (Tony Nolan).
234 080328 Rdataset: ensure column names valid R identifiers (Peter Christen)
233 080323 Review globals - continue moving into crv, but don't crv <- crv.
           Bug fix pmml.ksvm export - needs data.name
232 080319 Begin moving .VARs into crv variable to remove R CMD check NOTEs.
           Manage different working dirs for projects/data (Michael Fogliani).
           Bug fix: Ensure data textview is shown after load default audit.csv.
           Abbreviate generated variable names (otherwise they get too long).
231 080315 Bug fix: score CSV uses all columns, ignore as train (Ed Cox)
           If no CSV filename specified, load audit.csv (Tony Nolan)
           Add AsFactor and AsNumeric to transform (Greg Rose)
           Allow ByGroup transform to work without a group (Tony Nolan)
           Bug fix: check Select tab executed before Explore is attempted.
           Bug fix: drawTreeNodes incorrect with loss matrix (Rob Williams)
           Bug fix: kmeans fails with NA's so use na.omit.
           Bug fix: Transform tab should not expand to fill vertical space!
230 080304 Add export of PMML for nnet (zementis.com)
229 080304 Bug fix: risk plot - avoid integer overflow - use numeric.
228 080303 Bug fix: Logic for TARGET handling could result in two targets!
227 080302 Bug fix: Transform-Normalise-Centered had a syntax error.
226 080301 Bug fix: use of omitted in evaluate score and PrvOb (Ed Cox)
225 080229 Bug fix: checkbox wrongly checked Evaluate tab (Frank Lu)
224 080228 Add new By Group normalisation to Transform tab (Tony Nolan)
223 080204 Rene Koch: fix handling of ID, TARGET variables.
222 080202 Cleanup pmml.ksvm support. 
221 080202 Zementis: support for pmml from ksvm and binary logistic regression 
220 080202 Ed Cox: Pr v Ob plots into Eval tab.
219 080131 Add export of SVM into Rattle - not yet working in pmml.
218 080130 Zementis: addition of pmml export of ksvm.
217 080129 Ed Cox fix to hier cor plot labels. Fix generation of zip pkg.
216 080115 Add Help for Normalise.
215 080114 Export crs$glm using pmml.lm. Spelling fix in pmml.lm.
214 080106 Add new data source "Library" which lists all available R datasets
213 071220 Resync
212 071220 Update note on quantiles for binning
211 071212 Robust kmeans using multiple runs for clustering.
210 071212 Add subset sizes to Benfords legend.
209 071205 Explore->Distr->Cat: Add mosaic plots (Rohan Baxter)
208 071201 Transform Remap join categoricals for stratification.
207 071201 Benford: Add "digit/+/-" options (Eugene Dubossarsky)
206 071128 Bug fix: MSWindows useFancyQuotes=TRUE fails in textview (RGtk2)
205 071126 Add binning (Daniele Medri), log, indicators to Transform->Remap.
204 071125 Implement Transform->Cleanup to delete variables from the dataset.
           Export dataset from Data/Select tab. Depend on reshape. 
203 071124 Transform->Normalise implemented: scale, normalise, rank. (Ed Cox)
202 071124 Redesign transform tab; complete imputations (add mode + constant).
201 071117 Bug fix for RGtk2 (foreach). Implement mean and median imputation.
200 071113 Bug fix Benfords - didn't count negatives.
199 071113 Shade rpart tune options. Bug fix selectedForeach (Ignore/Input)
198 071112 Move to glade-3. Add basic tune functionality to rpart.
197 070929 Fix pmml.rsf to match new randomSurvivalForest version
196 070902 Export Kmeans as PMML (means) or CSV (clusters) (Enrico Vera-Cruz).
195 070902 Iterate kmeans clusters (Enrico Vera-Cruz).
194 070721 Brian Ripley suggestion re MS/Windows memory problem.
193 070709 Change ada stumps button to a checkbutton. Add pmml.lm.
192 070514 Fix generation of trees for rf.
191 070513 Improve RF rule generation, including R and VB code.
190 070506 Remove ARFF read/write and use from foreign if available.
189 070506 Cleanup: test missing packages, error messages, start removing arff
188 070417 Settings->Use Cairo Device.
187 070424 Add "?" as a missing value marker when reading CSV (Jason Ma).
186 070422 Ensure read.csv reads ",," as NA (Jason Ma).
185 070422 Check for undefined columns in evaluate and give message.
184 070422 Fix scoring where target/risk columns are not in the CSV file.
183 090422 Update various help texts.
182 070914 After a transform, keep variable roles primarily as they were.
181 070911 Define seq_len for pre 2.4.0 users of read.arff (Zefeng Dong).
180 070910 Use dev.print(win.print) under MS/Windows for printer selection.
179 070409 Clean up PMML and bump version to 1.1.0.
178 070408 Update status bar while "Applying model to data ..."
177 070408 Update status bar while "Building a model ..."
176 070408 When Build model uncheck other evaluate buttons (Michael Fogliani).
175 070408 Expose the commands to Save/Copy/Print in the Log tab.
175 070408 Implement plot annotations for boxplots.
174 070407 Clean up Cut/Save/Plot code and get working on MS/Windows.
173 070406 Trial dev.copy to clipboard for Plot window for MS/Windows.
172 070406 Print for Plot windows using modified dev.print (Tony Nolan).
171 070404 Add Copy button to Plot to save graphics to clipboard.
170 070401 For MS/Windows check if supplied csvname begins with "d:".
169 070326 Ensure paradigms go after Transform tab, not Explore tab.
168 070325 Error check and bug fix ada Continue and List/Draw buttons.
167 070325 Add Continue button to Ada to add trees to model (Peter Ricci)
166 070324 Add Degree option to Polynomial kernel for SVM.
165 070324 Inform ggobi of the name of the dataset.
164 070324 Add rectangles to show clusters in dendrogram.
163 070321 Add an Edit button to each of the Data options.
162 070321 Add a Data Entry option for manual data entry (Tony Nolan). 
161 070320 Include release date in About (Tony Nolan).
160 070319 Add kernel and class weights to SVM.
159 070319 UI changes: buttons have no relief; summary textview does not wrap.
158 070319 Help menu reflect new tab arrangement.
157 070317 Ensure "make check" is working - add to and correct docs.
156 070317 Question dialog improvements. Audit dataset generation improvement. 
155 070317 Place found string in summary search function further into the view.
154 070317 Move Sample to Select tab. Place Transform tab after Explore tab.
154 070317 Rename Variables tab to Select tab.
153 070317 Use a combobox to select type of imputation to perform.
152 070317 Allow a csvname to be supplied to rattle on startup (Tony Nolan).
151 070316 Add support for cairoPlots saving to SVG if RSvgDevice is available.
150 070316 Support reading of ARFF data files (Denny).
149 070310 Finish ada cleanup.
148 070308 Clean up ada as a sample template module.
147 070307 Allow using hclust centers to prime kmeans (Daniele Medri).
146 070303 Add a View Data button to view the loaded dataset.
145 070303 Add extra options for ada model (Frank Lu).
144 070303 Bug Fix - search in summary_textview now scrolls to selection.
143 070303 Use much more efficient hcluster with parallel processing.
142 070302 Add a Data Plot to KMeans clustering. Improve ODBC handling.
141 070228 Call gtkMainIteration() on file browser to update it immediately.
140 070227 Fix on a version scheme - 2.1.1 rather than 2.1.139 or 2.1.1-139.
139 070226 Bug fix for ODBC. Catch memory error for svm (Michael Fogliani)
138 070226 Add ability for user to enter an SQL query thru ODBC (Frank Lu)
137 070225 Bug fix for Basket Associations Freq Plot (Tom Arnold).
136 Add Limit and believeNRows to ODBC interface.
135 Add Find to Summary textview (Michael Fogliani)
134 Support Save of plot to wmf on MS/Windows.
133 Bug fix for Explore Dist Box plot - don't use summaryBy if no target
132 Release Version 2.2.0
132 Bug fix for eval of csv files - second time did not use "included"
131 Bug fixes - location of glade file.
130 Export listRPartRules (Frank Lu).
130 Move to using the Cairo device, with save/close support.
130 No longer save graphics with the use of the Export button.
129 Add support for saving most graphics.
128 Introduce pmml as an S3 function (Tobias Verbeke)
128 Implement export for Explore->Distr and Eval->Risk.
127 Ensure global variables are hidden (begin with ".") (Tobias Verbeke)
126 Bug fix non-sample ROCR plots (Tobias Verbeke)
125 Include randomSurvivalForest export of pmml (Udaya Kogalur)
124 Refine pmml implementation.
123 Bug fix - IGN vars in Distributions had no levels.
122 pmml now separately packaged and include pmml.kmeans
121 Bug fix - variable initialisation problem in pmml.print
120 Initial and basic rpart -> pmml: only >= and < supported. 
119 Begin migration of pmml code into own pmml package (Tobias Verbeke)
118 Bug Fix - missing internal.Rd docs.
117 Update some help documentation.
116 Add missing value summary (from mice) to Impute.
115 Initial imputation (Zero/Missing) (suggested by Michael Fogliani).
114 ROC plots of one model/testset also plot training (suggest Robert Williams)
113 Transform tab now uses radio buttons in preparation for imputation etc.
112 RF check if na.omit removes all entities (suggested by Michael Fogliani).
111 Activate proper model buttons on project load.
110 Report num entities with missing values in Summary (Michael Fogliani).
109 Market basket with arules (Suggested by Daniele Medri).
108 Initial nnet framework. Expose load/save commands in log. Bug fixes.
107 Save ada model when saving a project.
106 Bug fix project saving where crs$rpart fields were being overwritten.
105 Bug fix to not expose use of getSelectedVariables.
104 Add missing manual pages.
103 Ada options implemented. Support CSV without headers.
102 Add timestamps to modelling. Make Ada second modeller.
101 Add AdaBoost from the ada package (Suggested by Rohan Baxter)
100 Documentation updates.
99 RF can now display selected trees.
98 RPart shows rules on request. (Rules button)
97 Ada is now used for boosting.  (Suggested by Rohan Baxter)
96 Identify variable roles from names. (Suggested by Shaun Wicks)
94 Bug fixes.
93 Initial work on ada instead of gbm.  (Suggested by Rohan Baxter)
92 RPart default for the model tab, not All. (Suggested by Robert Williams)
91 Save/Load project set default directory. (Suggested by Michael Fogliani)
90 Add Togaware pointer to menu bar. (Suggested by Tony Nolan)
89 Bug: project load did not executeVariablesTab. (Noted by Frank Lu)
88 Include AUC in the Risk Charts. (Suggested by Frank Lu)
87 Include time taken in the log for model building.
86 Importance button for GBM. "All" as the default for model tab.
85 Replace RF checkbutton plots with buttons to plot.
84 Add an All radio button to Models tab (Suggested by Stuart Hamilton).
83 Progress rpart PMML export. (Suggested by Ray Lindsay)
82 RPart split out. On PMML export ask for file name.
81 Reset model check buttons on evaluate on a reset.
80 Work around library/source for arules inspect problems.
77 Initial Associations implementation.
76 Initial Associations interface.
75 Include Random Forest error rate plot.
74 Implement hclust stats and plot buttons.
73 Add a Dendrogram button to hclust.
72 Multiple risk charts now on one plot.
71 RPart does not display tree by default - use the button.
69 Add new options for KMeans. Generate random numbers for seeds.
68 Implement export of Log tab.
65 Move to multiple R src files and modularisation. Implement New button.
64 Initial implementation of paradigms.
63 Tools menu to navigate tabs (Anthony Nolan).
60 RF sampsize implemented (Stuart Hamilton)
59 Add varImpPlot to RF.
55 Support different summary options.
54 Fix plot colours in ROCR plots.
53 Add ggobi option to Explore tab (Marco Lo).
50 All packages loaded with quietly=TRUE.
48 All packages loaded with require.
47 Don't run examples in evaluateRisk.Rd (fails on CRAN MS/Windows).
45 Remove e1071's svm. Allow multiple model evaluation (Stuart Hamilton).
44 Allow target to be a weight variable (Liyin Xue).
43 Add a seed to the sampling for repeatable modelling (Liyin Zue).
42 Make ksvm the default SVM.
40 Add a biplot to principal components.
39 Remove seriation plot---too slow.
38 KMeans GUI now allows specifying k (Weiqiang Lin). 
37 Multiple plots per page.
29 Add a density plot to histograms (John Maindonald).
26 Add line plots for Benford's law (Tatiana Semenova).
23 Add Missing Value correlations (Ray Lindsay).
20 Add RData to file filter in loading Rdata files.
19 The default working directory follows latest openned file.
18 A Sample checkbutton has been added to the Explore tab.
14 Moved to hosting at google code.

Version 2.1 2006-10-23

General: Save/Load variable, model, and plot selections
Data: Load data from an ODBC source
Transform: Sample, Factorise, Imputation
Explore: Distribution plots (box, hist, cdf, Benford's Law, bar, dot)
Cluster: Visualise both kmeans and hierarchical clusters.
Model: Support SVM (kernlab)
Evaluate: Report AUC (area under the curve). Save scores to file.

Version 2.0 2006-06-10

General: R Package for inclusion in CRAN; GUI improvements; Bug fixes

Version 1.117 2006-05-27

General: Save/Load project functionality implemented.
Data: Allow other separators in CSV files!

Version 1.105 2006-05-06

General: Various code cleanup; More extensive help.
Features: Add a Weight Calculator for instance weights.
Sample: Sampling is now the default.
Summary: Include kurtosis and skewness measures.
Model: Add SVM (using kernlab) and Regression (using glm).

Version 1.77 2006-03-20

General: Reorder tabs more logically; Update Help; Implement Save; Export PMML for Kmeans
Features: Select roles for each feature
Explore: Add Hierarchical Feature Correlation; Add Principal Components
Cluster: Add Hierarchical, with dendrogram and seriation plots
Model: Add RandomForest and GLM
Evaluate: Clean up the interface for choosing a dataset
Log: Add timestamps and clean up a number of logs

Version 1.44 2006-03-05

Prepare: Introduce feature roles; allow multiple selection
Cluster: New tab with kmeans for a start
Model: Implement many RPart parameters
Evaluate: Load CSV/RDataset for training dataset.

Version 1.20 2006-02-27

Data: Link to any data frame in the current R session.
Explore: Add correlation plots.
Model: Add boosting using GBM.
Evaluate: Add Risk Charts
Log: Added a log tab to trace all actions for repeatability

Version 1.1 2006-02-15

Data: Load CSV files.
Explore: Data summary.
Prepare: Allow features to be ignored.
Select: Sample a percentage.
Model: Decision tree using rpart.
Evaluate: Confusion table and ROCR plots.