You can create expressions using an item from one of the following categories:

• Constants
• Field names
• Arithmetic operators
• Boolean operators
• Comparison operators

To create the constant 1997:

• Select Constants in the Category list.
• Double-click on <new constant> in the Value list.
• Type 1997 in the displayed entry field and press Enter.

You can edit any clause of the form Arg1=Arg2.

To edit a clause:

• Click the AND push button to add the default clause of the form Arg1=Arg2 to the Workarea container.

 
• Select a field name from the Value list and click on the Arg1 push button.

 
• Click, for example, the > push button to change the default operator.
   
• Create a new constant or select a constant from the Value list and click on the Arg2 push button.

 
• Optional: Click the AND or the OR push button and repeat the previous steps to use several clauses as filter records condition.
   
• Click the OK push button to add the expression to the Filter records condition field on the previous page.

The summary of your specification for the advanced pages and controls might look like this:   Control name Value Settings name Sample settings Comment Data collected during 1st half of 1997 Mining function Associations     Input data Sunset June transactions Comment Branch locations New York City Optimize mining run for Time Filter records condition Conditions selected Power options       Filter mask Selected Transaction ID Customer number Item ID Article Filter mask sales* Sort the input data on the values in the Transaction ID field before running this function False     Minimum support 5 Minimum confidence 75 Maximum rule length 3 Item constraints 5 selected Run the parallel  mode of this function Use this number of parallel processes: 4     Taxonomy name Sunset non-food Comment Different taxonomies in New York City, Washington D.C., and Cleveland     Results name Sunset June associations Comment Branch locations New York City If a result with this name exists, overwrite it False

 
After you have saved the settings for later use, this TaskGuide is closed. Remember to save the mining base from the main window before closing it.

You can use the settings individually or as part of a sequence.

You can use value mappings to define symmetric similarities for discrete fields. The value mapping contains two values and the similarity definition. During the clustering, both values use the specified similarity.

You can specify the similarity for each pair of possible values. One pair of possible values can only be used once. The similarity value for both values is the same. The similarity value must be between 0 and 1.

• 0 means completely different
• 1 means identical.

The summary of your specification for the advanced pages and controls might look like this:   Control name Value Settings name Sample settings Comment Data collected during 1st half of 1997 Mining function Demographic Clusters     Input data Quality food supermarket Comment May transactions Optimize mining run for Time Filter records condition Conditions selected Power options       Use mode Clustering mode Maximum passes 2 Maximum clusters 9 Accuracy 10 Similarity threshold 0,5     Active fields 6 selected Supplementary fields 3 selected Filter mask Customer     Field parameters  5 selected  Additional field parameters 5 selected Outlier treatment Treat outliers as missing values Similarity matrix 1 selected     Run the parallel mode of the function Use this number of parallel processes: 4 Additional parallel parameters Selected     Output fields 1 selected Cluster ID field name Cluster ID  Record score field name   Cluster ID field name choice 2   Record score field name choice 2   Confidence field name       Output data Quality food demographic clusters Comment  May transactions      Result name Quality food demographic clustering result Comment May transactions If a result with this name exists, overwrite it False

 
After you have saved the settings for later use, this TaskGuide is closed. Remember to save the mining base from the main window before closing it.

You can use the settings individually or as part of a sequence.

The summary of your specification for the advanced pages and controls might look like this:  Control name Value Settings name Sample settings Comment Data collected during 1st half of 1997 Mining function Neural Clusters     Input data Quality Food supermarket Comment May transactions Optimize mining run for Time Filter records condition Conditions selected Power options       Use mode Clustering mode Maximum passes 7 Maximum rows 5 Maximum columns 1     Active fields 6 selected Supplementary fields 3 selected Filter mask customer*     Outlier treatment Treat outliers as missing values     Run the parallel mode of the function Use this number of parallel processes: 4     Output fields 1 selected Cluster ID field name Cluster ID Record score field name   Cluster ID field name choice 2   Record score field name choice 2   Confidence field name       Output data Clusters Quality Food supermarket     Result name Result Quality food supermarket Comment May transactions If a result with this name exists, overwrite it False

 
After you have saved the settings for later use, this TaskGuide is closed. Remember to save the mining base from the main window before closing it.

You can use the settings individually or as part of a sequence.

The summary of your specification for the advanced pages and controls might look like this:   Control name Value Settings name Sample settings Comment Data collected during 1st half of 1997 Mining function Sequential Patterns     Input data Sunset retail June transactions Optimize the mining run for Time Filter records condition Conditions selected Power options       Transaction group field Customer ID Transaction field Date and purchase sequence number Item field Item ID Filter mask customer* Within each value in the Transaction group field, sort the input data on the values in the Transaction field False   Minimum support 5 Maximum pattern length 3 Item constraints Selected Run the parallel mode of this function Use this number of parallel processes: 4     Taxonomy name Sunset non-food Comment Different taxonomies in New York City, Washington D.C., and Cleveland     Results Sunset June sequential patterns Comment Branch locations New York City If a result with this name exists, overwrite it False

 
After you have saved the settings for later use, this TaskGuide is closed. Remember to save the mining base from the main window before closing it.

You can use the settings individually or as part of a sequence.

You can use the settings individually or as part of a sequence.

 

Summary: Neural Classification mining function

---

Use this page to verify the parameters you specified for the Neural Classification mining function, to change one of the parameters, to save the settings, or to start the mining function immediately.

The summary of your specification for the advanced pages and controls might look like this:   Control name Value Settings Sample settings Comment Data collected during 1st half of 1997 Mining function Neural Classification     Input data Insurance data Security First Comment Attributes: Age, salary, marital status Optimize mining run for Time Filter records condition "Risk class"=Good Power options       Use mode Training mode In-sample size 4 Out-sample size 2 Maximum number of passes 500 Accuracy 80 Error rate 20 Regardless of the mode, normalize the input data True     Input fields 4 selected Class label Risk class Filter mask *class*     Architecture determination Manual Hidden units 1 2 Hidden units 2 3 Hidden units 3 2 Parameter determination Manual Learn rate 0.2 Momentum 0.9     Run the parallel mode of the function 4 processes     Output fields 3 selected Class ID field name Risk class Confidence field name       Output data Output risk classes Security First Comment Customers who allowed their insurance to lapse     Result name Results Security First risk classes Comment Customers who allowed their insurance to lapse If a result with this name exists, overwrite it False

 
After you have saved the settings for later use, this TaskGuide is closed. Remember to save the mining base from the main window before closing it.

You can use the settings individually or as part of a sequence.

---

 
kêÜj
0¤ž^§½5Êp·5þq·5 pž./usr/lpp/IMiner/html/idmem5du.htm¤¤  

Summary: Tree Classification mining function

---

Use this page to verify the parameters you specified for the Tree Classification mining function, to change one of the parameters, to save the settings, or to start the mining function immediately.

The summary of your specification for the advanced pages and controls might look like this:   Control name Value Settings Sample settings Comment Data collected during 1st half of 1997 Mining function Tree Classification     Input data Insurance data Security First Comment Attributes: Age, salary, marital status Optimize mining run for Time Filter records condition "Risk class"=Good Power options None     Use mode Training mode Maximum tree depth 5 Maximum purity per internal node 90 Minimum records per internal node 7 Classify result None selected     Input fields 4 selected Class label Risk class Filter mask *class*     Field weights 5 selected     Run the parallel mode of the function Use 4 parallel processes     Output fields 3 selected Class ID field name Risk class Confidence field name       Output data Output Security First risk classes     Result Result Security First risk classes Comment Customers who allowed their insurance to lapse If a result with this name exists, overwrite it False

 
After you have saved the settings for later use, this TaskGuide is closed. Remember to save the mining base from the main window before closing it.

You can use the settings individually or as part of a sequence.

---

 
NT kêþf
1¤–^§½5Êp·5þq·5 r–./usr/lpp/IMiner/html/idmem6bu.htm¤¤  

Summary: Neural Prediction mining function

---

Use this page to verify the parameters you specified for the Neural Prediction mining function, to change one of the parameters, to save the settings, or to start the mining function immediately.

The summary of your specification for the advanced pages and controls might look like this:   Control name Value Settings Sample settings Comment Data collected during 1st half of 1997 Mining function Neural Prediction     Input data Insurance data Security First Comment Branch office Cleveland Optimize mining run for Time Filter records condition "Revenue">50.000 Power options       Use mode Training mode In-sample size 7 Out-sample size 4 Maximum number of passes 500 Forecast horizon 0 Window size 1 Average error 0.1 Regardless of mode, normalize the input data True     Active fields 2 selected Supplementary fields 1 selected Prediction field Revenue Filter mask None     List of values to predict Selected     Architecture determination Manual Hidden units 1 2 Hidden units 2 3 Hidden units 3 2 Parameter determination Manual Learn rate 0.2 Momentum 0.9 Generate quantiles True List of quantile limits 2, 10, 25, 50, 75, 90, 98 Run the parallel mode of the function On 4 processor nodes      Output fields 1 selected Predicted value field name Predicted value Lower quantile field name Lower quantile Upper quantile field name Upper quantile Output data Predicted sales revenue     Result name Results insurance data Security First Comment Branch office Cleveland only If a result with this name exists, overwrite it True

 
After you have saved the settings for later use, this TaskGuide is closed. Remember to save the mining base from the main window before closing it.

You can use the settings individually or as part of a sequence.

---

 
kêÌM
2¤­^§½5Êp·5þq·5 t­./usr/lpp/IMiner/html/idmem6cp.htm¤¤  

Output fields: RBF Prediction mining function

---

Use this page to specify whether to create output data.

Hints and tips are available for sorting the fields in the Available fields list.

You must specify a valid name for the following output fields:

• Predicted value
• Region ID

If you choose to generate quantiles, the output data also includes the following output fields, which field names you have to specify:

• Lower quantile
• Upper quantile

From the Available fields list, you can select additional output fields to be included in the output data.

Hints and tips

In the Available fields list, you can sort the fields by ascending or descending order. Click the right mouse button and move the cursor over the displayed selection field.

Filtering the available fields

You can filter particular data fields in the Available fields list by specifying a field name in the Filter mask entry field. You can also use wildcard characters.

Example:You might want to look for data fields related to customers. The filter condition looks like this:  Filter mask *customer* To redisplay all data fields in the Available fields list, delete *customer* and press Enter.

Moving fields between lists

To move fields from the Available fields list to the Output fields list:

• Click the >> push button to move all  fields.
• Select the fields to be moved and click the > push button to move the selected  fields.

Example: You might want to specify the following field names for the output fields:  Output fields Predicted value field name Region ID field name Lower quantile field name Upper quantile field name Customer number Predicted revenue Region ID Lower quantile Upper quantile

---

à kêo
3¤&_§½5Êp·5þq·5 u&./usr/lpp/IMiner/html/idmem6du.htm¤¤  

Summary: RBF Prediction mining function

---

Use this page to verify the parameters you specified for the RBF Prediction mining function, to change one of the parameters, to save the settings, or to start the mining function immediately.

The summary of your specification for the advanced pages and controls might look like this:   Control name Value Settings Sample settings Comment Data collected during 1st half of 1997 Mining function RBF Prediction     Input data Insurance data Security First Comment Branch office Cleveland only Optimize mining run for Time Filter records condition "Revenue">50.000 Power options       Use mode Training mode In-sample size 12 Out-sample size 12 Maximum number of passes 25 Maximum centers 500 Minimum region size 20 Minimum passes 5     Active fields 2 selected Supplementary fields 1 selected Prediction field Revenue Filter mask Customer     Generate quantiles True List of quantile limits 2, 10, 25, 50, 75, 90, 98     Output fields 1 selected Predicted value field name Predicted revenue Region ID field name Region ID Lower quantile field name Lower quantile Upper quantile field name Upper quantile Output data Output sales revenue     Result name Predicted sales revenue Comment   If a result with this name exists, overwrite it True

 
After you have saved the settings for later use, this TaskGuide is closed. Remember to save the mining base from the main window before closing it.

You can use the settings individually or as part of a sequence.

---

 
>< kêÐ~
4¤B _§½5Êp·5þq·5 wB ./usr/lpp/IMiner/html/idmep99o.htm¤¤  

Create settings from output data: Processing functions

---

This window allows you to create an Intelligent Miner data object from the database table or database view that a processing function uses as the output data. You can create a data object from the output data of all processing functions except for the Run SQL, Clean Up Data Sources, and Copy Records to File processing functions.  When you specify output data for a processing function, you either select an existing Intelligent Miner data object or specify the schema name and table or view name.  If you specified a schema name and table or view name for the output data, you can use this window to create an Intelligent Miner data object that refers to this database table or view.

You might enter the following parameters to create a data object from the database table containing the output data for the current processsing function:    Settings name Comment Use mode Filtered Customers Customer data filtered for purchases over $100 Read and write

---

/òX
kê$09use this window to create an Intelligent Miner data object that refers to this database table or view.    You might enter the following parameters to create a data object from the database table containing the output data for the current processsing function: