Introduction

When XLMiner™ runs a Prediction or Classification routine, it generates some values and coefficients which it uses to calculate the output. XLMiner 3.0 also generates a Stored model sheet which "stores" these values. It is generated along with the other outputs in that routine.
This Stored sheet can then be used to perform scoring on User's data set , using a utility XLM.calc. (Please write to support@xlminer.com if you want it.)

Here is some information on what values XLMiner™ stores in the Stored sheets :-

Regression (MLR, LR) : The coefficients of the regression equation.

Trees (Classification Tree, Regression Tree) : The Tree Rules for all the trees that were selected while generating the Stored model sheet.

Naive Bayes : Each variable, each class, each distinct value of that variable and the probability of getting that class.

K - Nearest Neighbors : The Training data as the model.

Neural Networks : All the weights between Input, Hidden and Output layers.

Discriminant Analysis : Discriminant coefficients for each class.

Once XLMiner™ generates this Stored Model sheet, the scoring is possible even if the user does not have XLMiner™. This utility can be used separately. Remember, the data set on which we want scoring done should have the same or more #variables and preferably of identical types as in the data set using which the Stored Model sheet is made. If the new data has less number of variables than the stored sheet then scoring from the stored sheet will not be possible.

1. Let us view the stored models for a few routines like MLR and Classification Tree.

   Run MLR as mentioned in the Example - Multiple Linear Regression.

   Along with the other output sheets, you can see one more worksheet, MLR_Stored_1. Open it.

   

   In the stored sheet, XLMiner™ has listed all the coefficients of the regression equation it generated while building the model. The stored sheet also displays the information about the variables it used and the parameters that we supplied. The Data Dictionary describes all the variables that the data set has and their types. The Mining Schema shows what variables and Parameters were used to generate the model. All this will be useful when we use this readymade model on a different data set. We will be able to use this stored sheet for a data set which has the same or more #variables and preferably of identical types. If the new data has less number of variables than the stored sheet, scoring will not be possible.

   If we are using this stored sheet to score from a new data, we will have to use a separate utility called XLM.calc. Please write to support@xlminer.com if you do not have it.

2. We can view one more stored sheet, that of the Classification Tree.

   Run the Classification Tree module as explained in Example - Classification Tree. See the stored sheet, CT_Stored_1.

   As in MLR stored sheet, XLMiner displays the Data dictionary and Mining Schema. This stored sheet also shows the various Tree Rules that were generated while building the model, the prior class probabilities and some other options.

   

   Using this we can get score on a new data set using XLM.calc.