RProp MLP - Data Preparation Regression - Compute

The input can be data with strings, integers and doubles. It should have a string variable called "Target" as a regression (numeric value) column.

This component prepares data to be used for the "RProp MLP Learner" node creating a neural network. It will:

• 01 Remove missing values

• 02 Convert categories to numbers

• convert integers to double (MLP seems to like that)

• 03 Normalize the double columns

• 04 Normalization of the Target column (later needed separately for De-Normalization)

• 05 provide a list of Columns being used

The conversions will be stored in PMML Files to be later used to prepare new data.

Input ports

1. Type: Table
   the original data
   the original data with String "Target" 0/1

Output ports

1. Type: Table
   Normalized WITH Target
   Normalized original data WITH Target
2. Type: PMML
   01 Missing Values (PMML)
   PMML setting the missing values
3. Type: PMML
   02 Category to Number (PMML)
   PMML converting the strings to numbers (double later)
4. Type: PMML
   03 Normalization (PMML)
   PMML for the Normalization
5. Type: PMML
   04 Normalization of Target (PMML)
   No description available
6. Type: Table
   05 Column Name
   Table Specs of the modified training data

External resources

• Hub: Heart Disease - Machine Learning Case - MLP - Parameter Optimization Loop
• Component: RProp MLP - Data Preparation Binary - Apply
• Node: RProp MLP Learner