Advanced settings for building models

Advanced settings for building models

In the "Model Creation" block, in addition to general settings for creating a scoring model, advanced settings are also available. To go to them, go to the "Advanced settings for creating a model" section.

The settings are intended to correct the algorithm for selecting important attributes for creating a model, as well as calculating the scoring score for each attribute value.

Illustration

In this section, you can change the following settings:
1. Marked as a good result – how exactly in the file that will be uploaded for creating the scoring model and testing, “Good” results will be marked in the first column of the file.
2. Marked as a bad result – how exactly in the file that will be uploaded for creating the scoring model and testing, “Bad” results will be marked in the first column of the file.
3. Factor – coefficient for the formation of the scoring score. The lower the value, the lower the final scoring score will be, but the difference between good and bad will also be smaller. Cannot be less than 1.
4. Offset – coefficient for the formation of the scoring score. The lower the value, the lower the final score will be. Cannot be less than 0.
5. Minimum needed total "Information Value" for Attribute – a metric that affects the selection of attributes without a conncat. The total information value is checked and, if it is less than the value specified here, the attribute is not taken into account for building the model.
6. Minimum needed average "Information Value" for Attribute – a metric that affects the selection of attributes without a conncat. The average information value among all attribute values is checked and if the average information value in the attribute is less than specified here, the attribute is not taken into account for building the model.
7. Max variants of value for Concated Attribute – after combining the attributes with each other for cross analysis, we will check the number of value options that turned out in the combined attribute, if there are more of them than specified here, then the attribute will not be taken to build the model. We do not recommend setting a value greater than 20 or 30 here, this is usually the maximum value so that the attribute does not distort the model.
8. Minimum needed total "Information Value" for Concated Attribute – a measure that influences the selection of combined attributes. The total information value is checked and, if it is less than the value specified here, the attribute is not taken into account for building the model.
9. Minimum needed average "Information Value" for Concated Attribute – a measure that influences the selection of combined attributes. The average information value among all attribute values is checked and if the average information value in the attribute is less than specified here, the attribute is not taken into account for building the model.