The Binning Explorer app supports the following tasks:
Binning Explorer enables you to import data by either
starting directly from the data stored in a MATLAB® table or by
loading an existing creditscorecard
object.
To start directly from data:
Place the credit scorecard data in your MATLAB workspace. The data must be in a MATLAB table, where each column of data can be any one of the following data types:
Numeric
Logical
Cell array of character vectors
Character array
Categorical
In addition, the table must contain a binary response variable.
Open Binning Explorer from the MATLAB toolstrip: On the Apps tab, under Computational Finance, click the app icon.
Select the data from the Step 1 pane of the Import Data window.
From the Step 2 pane, set the Variable
Type for each of the predictors, as needed. If the input
MATLAB table contains a column for weights
,
from the Step 2 pane, using the Variable
Type column, click the drop-down to select
Weights. If the data contains missing values,
from the Step 2 pane, set Bin missing
data: to Yes. For more information
on working with missing data, see Credit Scorecard Modeling with Missing Values.
From the Step 3 pane, select an initial binning algorithm and click Import Data. The bins are plotted and displayed for each predictor. By clicking an individual predictor plot, the details for that predictor plot display in the Bin Information and Predictor Information panes.
creditscorecard
ObjectTo start using an existing creditscorecard
object:
Place the creditscorecard
object in your MATLAB workspace. Create the creditscorecard
object either by using creditscorecard
or by
clicking Export in the Binning
Explorer to export and save a
creditscorecard
object to the MATLAB workspace.
Open Binning Explorer from the MATLAB toolstrip: On the Apps tab, under Computational Finance, click the app icon.
From Step 1 pane of the Import
Data window, select the creditscorecard
object.
From the Step 3 pane, select
a binning algorithm. When using an existing creditscorecard
object,
it is recommended to select the No Binning option.
To display the predictor plots, click Import Data.
The bins are plotted and displayed for each predictor. By clicking an individual predictor plot, the details for that predictor plot display in the Bin Information and Predictor Information panes.
creditscorecard
ObjectTo start Binning Explorer from the MATLAB command line:
Place the credit scorecard data or existing
creditscorecard
object in your MATLAB workspace.
At the MATLAB command prompt:
Enter binningExplorer(data)
or
binningExplorer(data,Name,Value)
to
open a table in the Binning Explorer app by
specifying a table (data
) as
input.
Enter binningExplorer(sc)
to open an
existing creditscorecard
object in the
Binning Explorer app by specifying a creditscorecard
object (sc
)
as input.
The bins are plotted and displayed for each predictor. By clicking an individual predictor plot, the details for that predictor plot display in the Bin Information and Predictor Information panes.
After you import data or a creditscorecard
object
into Binning Explorer, you can change the predictor type.
Click any predictor plot. The name of the selected predictor displays on the Binning Explorer toolstrip under Selected Predictor.
On the Binning Explorer toolstrip, the predictor type for the selected predictor displays under Predictor Type.
To change the predictor type, under Predictor Type,
select: Numeric
,
Categorical
, or
Ordinal
. The predictor plot is updated and
the details in the Bin Information and
Predictor Information panes are also
updated.
After you import data or a creditscorecard
object
into Binning Explorer, you can change the binning algorithm
for an individual predictor or for multiple predictors.
Click any predictor plot. The selected predictor plot displays with a blue outline.
Tip
When you select a predictor plot with the blue outline, a status message appears at the bottom of the Binning Explorer that displays the last binning information for that predictor. Use this information to determine which binning algorithm is most recently applied to an individual predictor plot.
On the Binning Explorer toolstrip, click Apply Monotone and select Monotone, Split, Merge, Equal Frequency, or Equal Width. The predictor plot is updated with a change of algorithm. The details in the Bin Information and Predictor Information panes are also updated.
To change the binning algorithm for multiple predictors, multiselect more than one predictor plot by using Ctrl + click to highlight each predictor plot with a blue outline.
Click Apply Monotone and select Monotone, Split, Merge, Equal Frequency, or Equal Width. All the selected predictor plots are updated for a change of algorithm.
After you import data or a creditscorecard
object into Binning
Explorer, you can change the binning algorithm options for an individual
predictor or for multiple predictors.
Click any predictor plot. The predictor plot displays with a blue outline.
Tip
When you select a predictor plot with the blue outline, a status message appears at the bottom of the Binning Explorer that displays the last binning information for that predictor. Use this information to determine which binning algorithm is most recently applied to an individual predictor plot.
On the Binning Explorer toolstrip, click Algorithm Options to open the Algorithm Options dialog box.
From the associated Algorithm options dialog box:
Monotone
For Trend, select one of the following:
Auto (default) — Automatically determines if the WOE trend is increasing or decreasing.
Increasing — Looks for an increasing WOE trend.
Decreasing — Looks for a decreasing WOE trend.
The value of Trend does not necessarily reflect that of the resulting WOE curve. The Trend option tells the algorithm to look for an increasing or decreasing trend, but the outcome might not show the desired trend. For example, the algorithm cannot find a decreasing trend when the data actually has an increasing WOE trend. For more information on the Trend option, see Monotone.
For Initial number of
bins, enter an initial number of bins
(default is 10
). The initial
number of bins must be an integer >
2
. Used for numeric predictors
only.
For Category Sorting, used for categorical predictors only, select one of the following:
Odds (default) — The categories are sorted by order of increasing values of odds, defined as the ratio of “Good” to “Bad” observations, for the given category.
Goods — The categories are sorted by order of increasing values of “Good.”
Bads — The categories are sorted by order of increasing values of “Bad.”
Totals — The categories are sorted by order of increasing values of the total number of observations (“Good” plus “Bad”).
None — No sorting is applied. The existing order of the categories is unchanged before applying the algorithm.
For more information, see Sort Categories
Split
For Measure, select one of the following: Gini (default), Chi2, InfoValue, or Entropy.
For Tolerance, specify a
tolerance value above which the gain in the information
value has to be for the split to be accepted. The
default is 1e-4
.
For Significance, only for the
Chi2 measure, specify a
significance level threshold for the chi-square
statistic, above which splitting happens. Values are in
the interval [0,1]
. Default is
0.9
(90% significance
level).
For Bin distribution, specify values for
MinBad —
Specifies the minimum number n
(n>=0
) of
Bads per bin. The default value is
1
, to avoid pure bins.
MaxBad —
Specifies the maximum number n
(n>=0
) of
Bads per bin. The default value is
Inf
.
MinGood —
Specifies the minimum number n
(n>=0
) of
Goods per bin. The default value is
1
, to avoid pure bins.
MaxGood —
Specifies the maximum number n
(n>=0
) of
Goods per bin. The default value is
Inf
.
MinCount —
Specifies the minimum number n
(n>=0
) of
observations per bin. The default value is
1
, to avoid empty bins.
MaxCount —
Specifies the maximum number n
(n>=0
) of
observations per bin. The default value is
Inf
.
MaxNumBins —
Specifies the maximum number n
(n>=2
) of
bins resulting from the splitting. The default
value is 5
.
For Initial number bins, specify
an integer that determines the number
(n >0) of bins that the predictor
is initially binned into before splitting. Valid for
numeric predictors only. Default is
50
.
For Category sorting, used for categorical predictors only, select a value:
Goods — The categories are sorted by order of increasing values of “Good.”
Bads — The categories are sorted by order of increasing values of “Bad.”
Odds — (default) The categories are sorted by order of increasing values of odds, defined as the ratio of “Good” to “Bad” observations, for the given category.
Totals — The categories are sorted by order of increasing values of total number of observations (“Good” plus “Bad”).
None — No
sorting is applied. The existing order of the
categories is unchanged before applying the
algorithm. (The existing order of the categories
can be seen in the category grouping optional
output from bininfo
.)
For more information, see Sort Categories
Merge
For Measure, select one of the following: Chi2 (default), Gini, InfoValue, or Entropy.
For Tolerance, specify the
minimum threshold below which merging happens for the
information value and entropy statistics. Valid values
are in the interval (0.1)
. Default is
1e-3
.
For Significance, specify the
significance level threshold for the chi-square
statistic, below which merging happens. Values are in
the interval [0,1]
. Default is
0.9
(90% significance
level).
For Bin distribution, specify the following:
MinNumBins —
Specifies the minimum number n
(n>=2
) of
bins that result from merging. The default value
is 2
.
MaxNumBins —
Specifies the maximum number n
(n>=2
) of
bins that result from merging. The default value
is 5
.
For Initial number of bins,
specify an integer that determines the number
(n >0) of bins that the predictor
is initially binned into before merging. Valid for
numeric predictors only. Default is
50
.
For Category sorting, used for categorical predictors only. Select a value:
Goods — The categories are sorted by order of increasing values of “Good.”
Bads — The categories are sorted by order of increasing values of “Bad.”
Odds — (default) The categories are sorted by order of increasing values of odds, defined as the ratio of “Good” to “Bad” observations, for the given category.
Totals — The categories are sorted by order of increasing values of total number of observations (“Good” plus “Bad”).
None — No
sorting is applied. The existing order of the
categories is unchanged before applying the
algorithm. (The existing order of the categories
can be seen in the category grouping optional
output from bininfo
.)
For more information, see Sort Categories
Equal Frequency
For Number of bins, enter the
number of bins. The default is 5
, and
the number of bins must be a positive number.
For Category Sorting, select one of the following:
Odds (default) — The categories are sorted by order of increasing values of odds, defined as the ratio of “Good” to “Bad” observations, for the given category.
Goods — The categories are sorted by order of increasing values of “Good.”
Bads — The categories are sorted by order of increasing values of “Bad.”
Totals — The categories are sorted by order of increasing values of the total number of observations (“Good” plus “Bad”).
None — No sorting is applied. The existing order of the categories is unchanged before applying the algorithm.
Note
You can use Category Sorting with categorical predictors only.
Equal Width
For Number of bins, enter the
number of bins. The default is 5
and
the number of bins must be a positive number.
For Category Sorting, select one of the following:
Odds (default) — The categories are sorted by order of increasing values of odds, defined as the ratio of “Good” to “Bad” observations, for the given category.
Goods — The categories are sorted by order of increasing values of “Good.”
Bads — The categories are sorted by order of increasing values of “Bad.”
Totals — The categories are sorted by order of increasing values of the total number of observations (“Good” plus “Bad”).
None — No sorting is applied. The existing order of the categories is unchanged before applying the algorithm.
Note
You can use Category Sorting with categorical predictors only.
Click OK. The predictor plot is updated with the change of algorithm options. The details in the Bin Information and Predictor Information panes are also updated.
To change the binning algorithm option for multiple predictors, multiselect more than one predictor plot by using Ctrl+ click to highlight each predictor plot with a blue outline.
On the Binning Explorer toolstrip, click Algorithm Options to open the Algorithm Options dialog box. Make your selection from the Algorithm Options dialog box and click OK. The selected predictor plots are updated for the change of algorithm.
After you import data or a creditscorecard
object
into Binning Explorer, you can split bins for a numeric
predictor.
Click any numeric predictor plot. The predictor plot displays with a blue outline.
On the Binning Explorer toolstrip, click Manual Binning to open the selected numeric predictor in a new tabbed window.
Click a bin to enable the Split button for that bin.
Note
The Split button is enabled when the data range of the selected bin has more than one value.
On the Binning Explorer toolstrip, the Edges text boxes display values for the edges of the selected bin. Click Split to open the Split dialog box.
Use the Number of bins control to split the selected bin into multiple bins. Click OK to complete the split operation.
The plot for the selected numeric predictor is updated with the new bin information. The details in the Bin Information and Predictor Information panes are also updated.
After you import data or a creditscorecard
object
into Binning Explorer, you can split bins for a categorical
predictor.
Click any categorical predictor plot. The predictor plot displays with a blue outline.
On the Binning Explorer toolstrip, click Manual Binning to open the selected categorical predictor in a new tabbed window.
Click a bin to enable the Split button for that bin.
Note
The Split button is enabled when the selected bin has more than one category in it.
Use the Number of bins control to split the selected bin into multiple bins.
Use the arrow controls on the Split dialog box to control the contents for each of the bins that you are splitting the selected bin into.
Click OK to complete the split operation.
The plot for the selected categorical predictor is updated with the new bin information. The details in the Bin Information and Predictor Information panes are also updated.
After you import data or a creditscorecard
object
into Binning Explorer, you can split or merge bins for
a predictor.
Click any predictor plot. The predictor plot displays with a blue outline.
On the Binning Explorer toolstrip, click Manual Binning to open the selected predictor in a new tabbed window.
Note
The Merge button is active only when more than one bin is selected. Only adjacent bins can be merged for numeric or ordinal predictors. Nonadjacent bins can be merged for categorical predictors.
To merge bins, select two or more bins for merging by using Ctrl + click to multiselect bins to display with blue outlines.
When performing a merge with a numeric predictor, the Edges text boxes on the Binning Explorer toolstrip display the values for the edges of the selected bins to merge.
Click Merge to complete the merge operation. The plot for the selected predictor is updated with the new bin information. The details in the Bin Information and Predictor Information panes are also updated.
After you import data or a creditscorecard
object
into Binning Explorer, you can change the bin boundaries
for a single predictor.
Click any predictor plot. The predictor plot displays with a blue outline.
On the Binning Explorer toolstrip, click Manual Binning. Click to select a specific bin where you want to change the bin dimensions. The selected bin displays with a blue outline.
On the Binning Explorer toolstrip, the Edges text boxes display values for the edges of the selected bin.
Edit the values in the Edges text boxes to change the selected bin’s dimensions.
Press Enter to complete the operation. The plot for the selected predictor is updated with the updated bin’s dimension information. The details in the Bin Information and Predictor Information panes are also updated.
After you import data or a creditscorecard
object
into Binning Explorer, you can change the algorithm applied
to one or more predictors and you can also redefine the number of
bins.
Click any predictor plot. The predictor plot displays with a blue outline.
Alternatively, select two or more predictors by using Ctrl + click to multiselect predictors to display with blue outlines.
On the Binning Explorer toolstrip, click Algorithm Options to open the Algorithm Options dialog box.
The Algorithm Options dialog box displays the options for the current binning algorithm. Depending on which is the current algorithm, you can change bin boundaries:
If your current algorithm for the selected bins is
EqualWidth
or
EqualFrequency
, enter a number in the
Number of bins box. Optionally, for
EqualWidth
and
EqualFrequency
options, under
Category Sorting, specify the type of
sorting.
If your current algorithm for the selected bins is
Monotone
, Split
, or
Merge
the default of
10
for Monotone
or
50
for Split
and
Merge
is used for the Initial
number of bins. Optionally, for
Monotone
, you can set values for
Trend and Category
Sorting.
Click OK to complete the operation. The plots for the selected predictors are updated with the new bin information. The details in the Bin Information and Predictor Information panes are also updated.
Binning Explorer has options for displaying predictor plots and plot options and the associated tables displayed in Bin Information.
From the Binning Explorer toolstrip item for Plot Options, select any of the following predictor plot options:
No labels (default)
Bin count
% Bin level
% Data level
% Total count
WOE curve
The selected label is applied to all predictor plots.
You can set the table display options for predictor information displayed in Bin Information.
From the Binning Explorer toolstrip item for Table Columns, select any of the following options:
Odds
WOE
InfoValue
Entropy
Chi2
Gini
Members (option is enabled for categorical predictors)
When selected, these options are applied to all predictors for the information displayed in Bin Information.
Binning Explorer enables you to export and save your
credit scorecard binning definitions to a creditscorecard
object.
Click Export and provide a creditscorecard
object
name. The creditscorecard
object is saved to the MATLAB workspace.
To reopen a previously saved creditscorecard
object,
click Import Data and select the creditscorecard
object
from the Step 1 pane of the Import Data window.
This topic shows some of the results when using Binning Explorer with credit scorecards that need troubleshooting. For details on the overall process of creating and developing credit scorecards, see Overview of Binning Explorer and Binning Explorer Case Study Example.
When you convert a numeric predictor with hundreds of values (for example, continuous data) to categorical data, the resulting data has hundreds of categories. The following example illustrates this scenario.
load CreditCardData
Open the Binning Explorer and select the numeric predictor AMBalance. From the Binning Explorer toolstrip, change the predictor type to Categorical.
Select Manual Binning on the Binning Explorer toolstrip and click Split. The Split dialog box displays as follows:
The predictor has too many categories to display properly.
Solution: If you have a categorical predictor with a large
number of categories, use the Algorithm Options to change
the binning algorithm for that predictor to Equal
Frequency, with the Number of bins set to
100
(or another smaller value). The Split dialog box then
displays properly.
When using the Binning Explorer, if you import data that has not been previously binned and you select No Binning from the Import Data window, the resulting plots might be distorted. For example, if you load the following data set into the MATLAB workspace and use Binning Explorer to import the data using No Binning, the following plot displays for the TmAtAddress predictor.
load CreditCardData
Solution: When you import data that has not been previously binned, select Monotone from the Import Data window instead. The following plot displays for the TmAtAddress predictor.