In any online platform, the new users will continue increasing. The previously discussed approach works well for the existing user. It is expensive to create a recommendation instance for every new user that's added. We cannot ignore the users that have been added to the system after the recommendation engine is made. To cope with situations that are similar to this, Apache Mahout has the ability of adding a temporary user to a data model. The general setup is as follows:
- Periodically recreate the whole recommendation using current data (for example, each day or hour, depending on how long it takes)
- Always check whether the user exists in the system before going for a recommendation
- If the user exists, then complete ...
