Using blocks

We will create our first block in our Rate project by importing the frequency model we built during the Risk tutorial. In the operations bar, click or use the keyboard shortcut M .

A screen will pop up that allows to select from importing a Risk model, importing a Demand model, or building a model from scratch. Start by clicking Risk, then navigating to the Risk model you created during the tutorial. Click NEXT in the bottom right.

Now we see the model tree and all of the tagged models from our Risk project. We will select our final model that included Geography and Classifications of both Geography and Vehicle. Then click ADD MODEL in the bottom right.

Before the model is imported, Akur8 will give us the ability to set the mapping of the model with Rate database. It will auto suggest a mapping based on the Risk model.

Since this Risk project was created with the same database as our Rate project, the mapping should line up well. However, this may not be the case when when the models trained on a different database. This tool allows you to verify the columns and levels match, the predictor type is correct, set up the Geography definition if it cannot be found.

Notice the column with the mapping score. This score is a rough measure of how well the Risk model matches to the Rate database. Half of the score is attributable to the column name matching, and the other half of the score is attributable to the levels matching. We can hover over the bars to get more detail about the scores for names and levels.

If a variable did not match up, we can click the change variable drop down for the row in question. This will give us more details on the matching score, the ability to map a different variable, and the ability to edit the coefficients. The latter is useful if the correct levels exist but have a new format from when the model was trained. For example, the model had a column with 1's and 0's, but the new database has TRUE and FALSE, we can easily edit the coefficients to apply them how we want without retraining the model.

Once we have confirmed the mapping, we click APPLY in the top right. We now have the model represented as a block in our Rate project!

When we left click to select our block, we'll notice a small menu pops up:

The sidebar is opened by default. You can open or close it using the chevron below the EXPLORE button.

The sidebar will vary depending on the type of block. Let us familiarize ourselves with the sidebar for model blocks.

Status: The status section shows us the various metrics and visualizations that this block will produce such as prediction histograms and loss related metrics. The icons on the left denote the status of these items:

Notice that the status for Target visualisation and Loss-related metrics are Inactive. We can change these to Ready by defining the parameters.

We see that the status for each item is now ready for computation as denoted by. We are ready to compute our block. At the bottom of the sidebar, click the COMPUTE button. We will discuss computing in greater detail once our project is more complex, but for now we can proceed. After a moment we should see that the status for each item has changed to.

Before we explore our frequency model, let's first build a Model from scratch for Severity. In the operations bar, clickor use the keyboard shortcut m. This time, select Build.

This process will allow us to create a multiplicative log-link GLM from scratch without leaving the Rate module. Here we have the option to set the initial average value, which is defaulted to 1. We can change the average value of this block later so keep this default value for now. Set the name as "PD Severity" and then click ADD MODEL in the bottom right to continue.

A new screen pops up that looks similar to the edit coefficients feature in Risk and allows us to start programming our model. Every variable in the dataset is available to us. When a variable is selected, we see the factor curve displayed against the exposures for each level. We also have an ACTIONS pane where we can visually modify selected points as well as a VALUES pane where we can paste in values. At the top, we also have the ability to edit the name and fix the average value.

Search for the variable vehicle_value_at_purchase and navigate to the VALUES pane. Copy the values from the below values and paste directly into the Coefficient % column.

Now we have the coefficients for our one variable severity model programmed. Note, if you wanted to define more variables you would simply repeat the steps above. Click APPLY in the top right and the block will appear in our rate project.

Now that the severity block is created, follow the steps below to define this block's parameters.

Notice the difference in shades between the two blocks. Each type of block has a designated color. In this case, model blocks are blue. The shade denotes if it has been computed or not. The darker shade for our Frequency model denotes that is has been computed, while the severity model is denoted by a lighter shade since it has not. Lets compute the severity block now.

We can explore the model blocks now that they have been computed. Select the Frequency block and click EXPLORE.

On the left of this window, we have a bar that displays the other blocks that can be explored. This allows us to quickly toggle across blocks without having jump back to the main workflow.

On the right, we have 3 panes:

Now that we have both frequency and severity blocks, we can combine them into a pure premium using the product operation.

After creating the PD Pure Premium block, notice the following:

Lets set the parameters and compute this block:

Now that this block is computed you can explore it in the same way that you explored the frequency and severity blocks. Notice that the overview and variable tabs are still populated. When multiplying two multiplicative GLM blocks together the resulting block will also be a multiplicative GLM. This feature will be extremely helpful when making selections and defining scenarios later.