It is worth mentioning that the first model developed in the first project cannot be persisted since it is just a few lines of code for computing movie similarity. It also has another limitation that we did not cover earlier. It can compute the similarity between two movies, but what about more than two movies? Frankly speaking, a model like the first one would rarely be deployed for a real-life movie. So let's focus on the model-based recommendation engine instead.
Although ratings from users will keep coming, still it might be worth it to store the current one. Therefore, we also want to persist our current base model for later use in order to save time when starting up the server. The idea is to use the current model for real-time movie recommendations.
Nevertheless, we might also save time if we persist some of the RDDs we have generated, especially those that took longer to process. The following line saves our trained ALS model (see the MovieRecommendation...