In case of the p-value generated by two or more Algorithms has to be combined, it’s useful to transform the p-values to the same scale before the combination. This can be done by using the “Predictor Normalizer” block after the Predictor.
This Node can be used to improve the variability of the Recommendations by filtering out too many items with the same Attributes. E.g. the below example enables only having two items with the same MaterialGroup value.
Sometimes the item property which should be used e.g. for filtering is not available in the database, but it can be calculated somehow. For this use-case dynamic item properties can be defined. After calculated, dynamic properties can be used similarly to standard item properties, but only in Nodes connected after the “Dynamic Property” Node. As it’s visible on the figure below, dynamic properties can be defined in several different ways.
Also, it can be used to save the actual p-value as item property for later use.
Cache Node is for performance optimization to avoid making time-consuming calculations too frequently. It does not change anything on the input just saves it, and if it’s expected to provide items on its output with the same context parameters, then it returns the saved list and the Nodes before the cache blocks are not executed.
Using Cache Node is recommended only for advanced users, because it makes debugging more difficult.
Updated about 3 years ago