Recommendation engines require large amounts of training data in order to do a good job, which is why they're often relegated to big data projects. However, to build a recommendation engine, we must first get the required data into memory and, due to the size of the data, must do so in a memory-safe and efficient way. Luckily, Python has all of the tools to get the job done, and this recipe shows you how.
You will need to have the appropriate movie lens dataset downloaded, as specified in the preceding recipe. If you skipped the setup in Chapter 1, Preparing Your Data Science Environment, you will need to go back and ensure that you have NumPy correctly installed.