Selecting the Data to Analyze


In order to run Cognalysis MultiRate, the data must be in Microsoft SQL Server. If the data is not already in SQL, MultiRate's Import Tool may be used to import the data into SQL.


SQL Server

The first item to select is the name of the SQL Server. A listing of the currently-available SQL Servers is pre-populated in the SQL Server drop-down box. Select the server that contains the data to be analyzed.



Once a server is selected, the Database drop-down box will populate with the databases that are currently attached to the server. Select the correct database for the analysis.


Combined Exposure and Target Table

Once the SQL Server and database have been selected, select the table that contains the Training Data to be analyzed. Test Data may be selected as well if desired.


Exposure and target information must be contained within one table. If the exposure and target information reside in separate tables, they must be joined into one table prior to running the MultiRate analysis.  If test data is used, the training and test data may be in different tables, but the the two tables must be formatted identically. If there is no separate test table, a Filter may be used to divide the data into two sets, one for training and one for test; adding a Random Characteristic can be useful for this.


The Filter may also be used to filter the data so that a subset of the data is used for the analysis rather than the entire dataset. For example, if the user only wants to analyze policies with nonzero earned premium, they could include a "where EP > 0" clause to exclude the zero-premium policies rather than create a whole new separate dataset without them.


The data does not need to be pre-aggregated into groupings or bins in order to run the analysis. Cognalysis MultiRate will automatically bin the data, and the bin breakpoints can be overridden by the user.