Evaluate
From Clariopedia
Contents |
Overview
The clario® node Evaluate is used to generate model documentation, including a model description, a model profile and related charts. Evaluate will work with either Logistic or Linear regression.
Usage
Input Stream
There are two ways to run Evaluate. These are outlined below as Method a and Method b.
Method a. Two input data streams are connected to the evaluate node: one dataset created as a result of a logistic or linear node (this dataset contains the model information only), and one dataset containing the actual modeling data (dependent and predictor attributes).
Connect the linear or logistic node to the upper connector. Connect the data stream to the lower connector.
Method b. Only one input data stream is connected to the evaluate node: a dataset that already contains a model score, which you wish to use to generate the model description, profile and charts.
Connect the data stream to the lower connector.
Configuration
The Evaluate node has only one configuration face.
Configuration Face
NOTES: To efficiently find attribute names, begin typing an attribute name in the text box directly above Available Attributes. You will then be directed to the attribute(s) beginning with the letter(s) you type. To select multiple attributes at once, either use [Ctrl]+click to select multiple, one at a time, [Shift]+down arrow to select multiple in order of appearance, or use [Shift]+click to select the beginning and the ending attribute which will select all attributes. To de-select an attribute click on the attribute in the Selected Attributes box and drag and drop into the Available Attributes box. Attributes in the Selected Attributes list can be re-ordered by clicking and holding on an attribute and dragging it to the desired position within the Selected Attributes box.
Method a – two inputs
Below the list boxes, select your dependent attribute from the drop down box. Next, type in the attribute name you want from the Model Score, which will be calculated based on your top connector model equation (linear or logistic). Next, select the # of segments you want for the model documentation, the profile and the charts. The valid range is 1-20 segments. Finally, select the weight attribute, if any, from the drop down box. If you do not have a weight attribute, leave the drop down box blank.
Method b – one input
Below the list boxes, select your model score attribute from the drop down box. Next, select the # of segments you want for the model documentation, the profile and the charts. The valid range is 1-20 segments. Finally, select the weight attribute, if any, from the drop down box. If you do not have a weight attribute, select the blank box.
Field Definitions
Because Evaluate gives you the ability to name attributes, a specific number of keys are valid. These valid keys are: A-Z, a-z, 0-9, "-", "_". If invalid keys are pressed when the text box is open, nothing will appear.
- Model streams are optional
- If a model stream (e.g. linear or logistic) is included, then the model stream node must be connected to the top connector.
- "Model Score Name" cannot be <NULL>.
- If no model stream is attached, the attributes from the data stream (bottom connector) are the data provider for the "Model Attributes" drop-down.
- At least 1 attribute must be in the "Selected Attributes" column.
- The default value for "Weight Attribute" is <NULL>.
- The selected "Weight Attribute" from the drop down list must consist of integer values.
- The range for the "Segments" spinner is 1-20.
- The default value for the "Segments" spinner is 10.
- An attribute cannot be a "Model Score Attribute" and a "Dependent Attribute" concurrently.
- The default value for "Dependent Attribute" is <NULL>.
- A workflow cannot be submitted with a value of <NULL> for the "Dependent Attribute."
Results
There is one results face for the Evaluate node.
Model Profile
This table profiles the results of your model, split into the number of segments you specified (between 1 and 20), using model score. The model scores are ordered low to high, and an attempt is made to rank all rows in your data into equally-sized segments. The profile then reports the number of rows for each segment, along with the means of the
Charts
This results table is actually a graphing tool, which you can use to graph any of the data in the Model Profile. The x-axis represents model segment, and you can graph 1 or 2 attributes on the y-axis. Just select the attribute you want to graph in orange in the drop down box next to the orange square. Then, if you want to graph a second attribute, select the attribute you want to graph in green in the drop down box next to the green square. The orange y-axis values are on the left side of the graph, and the green y-axis values are on the right side of the graph. One common graph will be to compare the Dependent Attribute and Predicted Score by segment.
Model
This shows the actual model equation using ‘pseudo code’ in the top box. It also shows each attribute in the model, along with the coefficient and score contribution (computed using the absolute value of the standardized estimate). Note that the score contribution is a percentage value, and all values add to 100. You can easily export the model description and coefficients to a spreadsheet by clicking on the Export to Spreadsheet button at the upper right.
IMPORTANT: When you are using Method b, the Model results screen will not be present, as you already have the model score as an attribute on your input data stream.
Output Stream
Two of the Evaluate results tables, Model and Model Profile, can be exported into Excel. There is no data file output from Evaluate, as it is a terminal node.
Video Demonstration
References
None.
