COMP3430 DW: Lab 2 Transform/Wrangle

[gjwgit]

See https://datamining.togaware.com/survivor/Transforming_Data.html

Specification

Primary GUI Tasks

• [x] There will be a collection of horizontal tabs for each transform feature: RESCALE, IMPUTE, RECODE, CLEANUP.

CONFIG

• [x] The CONFIG widget for each feature will have the same options in the Rattle screenshots below.
• [x] The chosen config options determine which R script to run for each chosen variable
• [x] Need to check compatibility of the script with the variable type (numeric versus categoric)

DISPLAY

• [ ] Each feature panel in the Wrangle tab needs a display.dart to implement a VariableChooser instead of Pages
• [ ] The VariableChooser will display a line for each variable and a click will select the variable
• [ ] VariableChooser line will include a summary like in the screenshots below.
• [ ] Multiple variables can be chosen.
• [ ] Model the VariableChooser look on the old Rattle perhaps with left hand side check buttons

BUILD

• [x] The BUILD button will perform the transformation by running the appropriate R script
• [x] Business rules include matching selected variable data type to R script data type
• [x] The R scripts update the ds with new variables named like XXX_SELECT_VAR as below.
• [x] The VariableChooser view of the dataset is updated to allow further transforms
• [x] The Dataset tab shows the new dataset
• [x] All state variables are updated appropriately
• [x] The transformed variable becomes IGNORE by default
• [x] The newly created variable becomes INPUT by default

R Scripts

• [x] There will be different R scripts for each operation
• [x] Scripts will be named as in transform_rescale_recenter_numeric.R.
• [x] The scripts will include parameters to be substituted:
  • [x] SELECTEDVAR (the original variable name used in all transform*.R scripts);
  • [x] MISSING_NUMERIC_VALUE; (for transform_impute_constant_numeric.R),
  • [x] MISSING_CATEGORIC_VALUE (for transform_impute_constant_categoric.R)

Rescale

Business Rules

• [x] Only NUMERIC variables can be re-scaled.

When the rescale feature is chosen the following two groups of rescaling functions are available with only one chosen at any time (radio buttons?):

*Normalise

• [x] RECENTER: name the new variable RRC_old_name
• [x] MEDIAN/MAD: RMD_old_name
• [x] NATURAL LOG: RLG_old_name
• [x] LOG10: R10_old_name

Order

• [ ] Rank
• [ ] Interval (requires a Groups - integer)

Impute

Business Rules

• [x] The variable must have missing values before calling the impute scripts. Otherwise it's a noop anyhow

For IMPUTE, for CATEGORIC variable

• [x] 'missing':
• [x] mode: IMO_old_name
• [x] constant: ICN_old_name

For IMPUTE, for NUMERIC variable

• [x] zero: IZR_old_name
• [x] mean: IMN_old_name
• [x] median: IMD_old_name
• [x] mode: IMO_old_name
• [x] constant ICN_old_name

[gjwgit]

Lab 2 is in Week 4 12 August

[gjwgit]

RESCALE and IMPUTE IMPLEMENTED