The hardware and bandwidth for this mirror is donated by METANET, the Webhosting and Full Service-Cloud Provider.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]metanet.ch.

modelStudio 3.1.2

• added new parameter to modelStudio(): open_plots = c("fi"), which is a vector listing plots to be initially opened (and on which positions) (#112)
• fixed future warning with DALEX::loss_default() since DALEX >=2.5.0

modelStudio 3.1.0

• changed y-axis variable labels in SV to the same as in BD
• added new parameter to modelStudio(): max_features_fi = max_features, which allows displaying a distinctive number of features in FI plot (other than in BD and SV)
• added new options to ms_options(): **_axis_title, which allow changing plot-specific axis title (default varies)

modelStudio 3.0.0

• BREAKING CHANGES:
  • this version requires R >=3.6, DALEX >=2.2.1, ingredients >=2.2.0 and iBreakDown >=2.0.1
  • the deprecated alias modelStudioOptions() is removed from this version of the package; after being deprecated for over a year since v1.1.0. Use the recommended ms_options() instead.
  • added new parameter to modelStudio(): N_sv = 3*N, which by default decreases the number of observations used for the calculation of Shapley Values (rows in data)
  • margin_left = NULL by default and it is adjusted based on the length of variable names
  • the first plot opened in the dashboard is now FI instead of BD by default
• added the verbose parameter to modelStudio() as an alias to show_info (#101)
• added new ms_merge_observations() function that merges local explanation of observations from multiple modelStudio objects (#102)

modelStudio 2.1.2

• fixed an error in modelStudio() when data had only one variable (#99)

modelStudio 2.1.1

• fix CRAN checks

modelStudio 2.1.0

• DEFAULTS CHANGES: if new_observation = NULL then choose new_observation_n = 3 observations, evenly spread by the order of y_hat. This shall always include the observations, which ids are which.min(y_hat) and which.max(y_hat). Additionally, improve the observation dropdown text in dashboard. (#94)
• updated the progress printing
• this version requires DALEX v2.0.1
• added new options to ms_options: ms_subtitle, ms_margin_top and ms_margin_bottom
• added new parameters to modelStudio(): N_fi = 10*N and B_fi = B
• added new license parameter to modelStudio() which allows to specify the connection for readLines() (e.g. 'LICENSE') which will add file contents into the HTML output as a comment

modelStudio 2.0.0

• this version requires DALEX v2.0, ingredients v2.0 and iBreakDown v1.3.1
• The dashboard gathers useful, but not sensitive, information about how it is being used (e.g. computation length, package version, dashboard dimensions). This is for the development purposes only and can be blocked by setting telemetry to FALSE.
• add support for modelStudio in Shiny (#77) using new widget_id argument
• modelStudio now works with NA in data (#71)
• CP, PD and AD plots are now calculated with variable_splits_type='uniform' and CP plots are now calculated with variable_splits_with_obs=TRUE (#74)
• By default the loss_function in FI is now different for each explainer$model_info$type (#73)
• fixed a bug where passing additional parameters in ... would cause an error
• added a max_vars alias for the max_features parameter
• added median line to the boxplots in FI and SV plots, added boxplots to TV categorical plots (regression)
• TV plot uses boxplots and barplot when the target y has only two unique values (classification) (#76)
• added more checks for input
• added the Residuals vs Feature plot (RV) (#84)
• added model performance measures to the footnote #(85)

modelStudio 1.2.0

• remove redundant documentation resources so that the package weights less
• add a second dropdown list for variable change
• fix check class warning
• add stringsAsFactors=TRUE where data.frame is used

modelStudio 1.1.0

• rename modelStudioOptions() to ms_options()
• add new ms_update_options() function that updates the options of a modelStudio object
• add new ms_update_observations() function that updates the observations of a modelStudio object
• lower B default value from 15 to 10 and N default value from 400 to 300
• feature_importance is now calculated on 10*N sampled rows from the data
• use ranger instead of randomForest everywhere
• remove unnecessary imports, update the documentation
• added auto_unbox = TRUE to jsonlite::toJSON and changed the .js code to comply
• add new class "modelStudio" to the modelStudio() output

modelStudio 1.0.2

• fix FD plot on matrix-like data
• center modelStudio position in HTML
• modelStudio now renders properly from .Rmd to .html

modelStudio 1.0.1

• fix devel-fedora tests for cran

modelStudio 1.0.0

• stable release after fixing minor issues
• comply with R v4.0 changes
• add support for matrix-like data with xgboost working example
• boxplot whiskers end in max(min, q1 - 1.5*iqr) and min(max, q3 + 1.5*iqr)
• upgrade show_info with progress package

modelStudio 0.3.0

• modelStudio() now only works on explainer class object made with DALEX::explain()
• this version requires iBreakDown v1.0.0 and ingredients v1.0.0
• change cummulative to cumulative in code (#49)
• change dependency to dependence in code (#52)
• update package title and description
• change LICENSE to GPL-3 (#55)
• add boxplots to SV plot (#50)
• add eda argument to modelStudio()
• add show_boxplot argument to modelStudioOptions()

modelStudio 0.2.1

• fix TV plot (X had columns sorted while y was the same)
• add ms_title argument to modelStudioOptions() (#46)
• modelStudio footer is generated faster

modelStudio 0.2.0

• new plot: Target vs Feature [EDA] (#38)
• new plot: Average Target vs Feature [EDA] (#41)
• add boxplots to FI plot ingredients/72
• add new_observation_y argument to modelStudio() (#39)
• pass ... to prepare_* functions (e.g. allows to round numbers)
• add margin_ytitle argument to modelStudioOptions()
• by default: observations to calculate local explanations are taken at random from the data (#25)
• by default: first plot is selected as BD and second is clicked (#37)
• nicer histogram when few unique values
• all categorical plots now have the same bar order ingredients/82
• try-catch blocks added - errors in ingredients or iBreakDown functions are now treated as warnings and do not stop the modelStudio computation (#35)
• show_info adds messages saying what is currently calculated (#40)
• add spellcheck to tests (#36)
• Travis-CI now checks OSX

modelStudio 0.1.9

• this version requires DALEX v0.4.9 and ingredients v0.4.0

modelStudio 0.1.8

• lower B default value from 25 to 15, N default value from 500 to 400

modelStudio 0.1.7

• fix tests for CRAN

modelStudio 0.1.6

Many minor changes stated in #20, most notably: * rename x parameter to object in modelStudio() * rename getOptions() to modelStudioOptions() * add viewer parameter to modelStudio() * add suppressWarnings(ingredients::describe) * upgrade documentation, examples and vignette

modelStudio 0.1.5

• add description to all plots besides AD and FD
• add footer to modelStudio
• change only_numerical to variable_type in ingredients functions

modelStudio 0.1.4

• add support for parallel computation with parallelMap
• more modelStudio customization with options parameter
• add getOptions() function
• remove plot subtitles by default

modelStudio 0.1.3

• remove file paths of dependencies from html file
• add animations to line plots
• change .js dependencies to min.js

modelStudio 0.1.2

• major .js code refactoring
• add proper exit plot buttons
• import Fira Sans font
• modelStudio does not reload on resize
• rewrite d3-tip code, fix placement, add pointer

modelStudio 0.1.1

• add demo, cheatsheet, animated instructions, more tests and examples
• description won’t show if there are less than 4 features in the model

modelStudio 0.1.0

• first stable version of the modelStudio package
• modelStudio() function implemented

These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.