Build Your First Structured Model

In this tutorial, you will build and run a structured model with one input. By the end, one record with an LTV of 0.82 will produce a verified score of 1.74.

What you will build

The model uses no normalization, so its output is the raw regression value:

score = 0.10 + (2.00 × ltv)

Input	Value	Expected output	Value
ltv	0.82	score	1.74

This deliberately uses a direct Dictionary Reference. Dynamic fields and scenario factors come later in the use cases.

Before you begin

You need:

• a running VOR Stream environment and permission to create a model
• a Model Type and LOB available in the Models editor
• a playpen in which you can edit src/dictionary.csv and src/tables.csv, add a stream file and run vor commands

The process selects the model by name, so this tutorial does not require a Framework or scenario. For background, see Creating Model Nodes.

1. Prepare a runnable record

Ensure these variables exist in the playpen's src/dictionary.csv:

name,type,descr,arraylen,group,genformat
ltv,num,Loan-to-value ratio,0,portfolio,
score,num,Structured model score,0,model_output,

If the file already has a header, its column order may differ. Add only missing variables, placing each value under the matching header. An existing ltv or score row must have type num; keep its other metadata. If the name already has another type, use a different name consistently throughout the tutorial.

Add queue definitions using the layout already present in the playpen's src/tables.csv:

tablename,varname layoutname layout

tablename,varname,type,descr,arraylen,inherit,groupkey
starter_input,,,Structured model tutorial input
starter_input,ltv,num
starter_output,,,Structured model tutorial output,,starter_input
starter_output,score,num

name,type,descr,inherit,groupkey
starter_input,,Structured model tutorial input,,
ltv
starter_output,,Structured model tutorial output,starter_input,
score

Do not mix the two layouts. The name layout requires name as its first header because the parser recognizes each field by its single-column row. Insert each field directly below its table row and before the next table. The tablename,varname layout supports other header orders; place each value under the matching header. Add only missing definitions, and confirm starter_input contains ltv, starter_output contains score, both fields resolve to type num, and starter_output inherits starter_input. If an existing definition is incompatible, use a different name consistently throughout the tutorial.

Create structured_model_tutorial.strm in the playpen root:

name structured_model_tutorial

in input.csv -> starter_input

model (starter_input)(starter_output)
    model_name="Starter LTV Score"

out starter_output -> output.csv

Create input/input.csv:

ltv
0.82

Register the variables before the tables, because table registration resolves dictionary references. Then generate the queue and process code:

vor update dictionary --file src/dictionary.csv
vor update tables --file src/tables.csv
vor create queue --data src/tables.csv
vor create process structured_model_tutorial.strm

The score output column is essential. A model result whose name matches no output queue column is dropped even when the run succeeds.

2. Create the model

1. Open the Models tab.
2. Click the Model icon.
3. Enter Starter LTV Score as the Model Name.
4. Select the Model Type and LOB available for this tutorial in your environment.
5. Select Structured above the editor.

The Dynamic Fields panel appears above Structured Regression. Leave it empty for this model: ltv already exists as a dictionary column and can feed a predictor directly.

3. Configure the regression

Enter these values in Structured Regression:

Setting	Value
Dependent Variable (Y)	score
Function	Regression
Normalization	None
Intercept (β₀)	0.10

The dependent variable names the result. It must match the score column you declared on starter_output.

4. Add the predictor

Click Add Predictor, then enter:

Predictor field	Value
Source	ltv
Coefficient (β)	2.00
Exp	1

The formula preview should now have this shape:

Y = 0.1 + 2 × ltv

Notice that the preview combines the intercept, coefficient and predictor. If it does not, compare each value with the two configuration tables before continuing.

5. Save the model

Click the icon, enter the save comment and confirm the save. Starter LTV Score should appear in the Models list.

Saving verifies the model definition. Running it verifies that the model, process and output queue agree.

6. Run the model

From the playpen root, run:

vor run structured_model_tutorial

The process reads the one input record, selects Starter LTV Score by the model_name in the stream file and writes output/output.csv.

7. Verify the result

Open output/output.csv and find the record with ltv = 0.82. Its score should be 1.74:

0.10 + (2.00 × 0.82) = 1.74

ltv	score
0.82	1.74

If the run succeeds but score is missing, confirm that:

1. the model's Dependent Variable (Y) is named score
2. starter_output declares a score column
3. queue code was regenerated after adding the column

See What a model node writes to its output queues for the complete output rule.

What you have learned

You have followed the complete structured-model loop:

1. declare the model's input and output fields
2. configure a regression from a dictionary predictor
3. use the formula preview to check the equation
4. save, run and verify a known result

Next, use the structured model use cases to add transformations, lookups, scenario factors and chained models. Return to the Structured Models overview when you need the field dependency rules.