Recommender Systems – Predict What Movies a User Would Like to Watch

Recommender systems, as the name suggests, are solutions that are designed to provide recommendations to users based on various parameters, such as past behavior, item similarity, or even user demographics. These systems are used in a range of applications, such as for suggesting videos on YouTube, movies on Netflix, or products on Amazon.

The primary goal of recommender systems is to personalize online user experiences to drive business outcomes such as higher user engagement and increased revenues. As the amount of available content and choices increases, personalized recommendations become crucial for enhancing user experience and ensuring that the customers don’t get overwhelmed by the available options.

In this chapter, we will cover the following topics:

• Overview of the different types of recommender systems
• Deploying a movie recommender system on Vertex AI

First, we’...

Different types of recommender systems

In this section, we’ll delve into the diverse types of recommendation engines, shedding light on their methodologies and the unique advantages each brings to the table:

• Collaborative filtering:

  This approach is based on the idea that users who have agreed in the past will agree in the future about their preference for certain items. As shown in the following figure, the model tries to find similar users by looking at their viewing/reading and recommends the content viewed by one user to other, similar users:

Figure 15.1 – Collaborative filtering

• Content-based filtering:

  This method uses item attributes to recommend additional items similar to what the user likes, based on their previous actions or explicit feedback.

  For example, if a user has shown a preference for movies directed by Christopher Nolan, the system will rank the movies that were directed by him higher when making recommendations...

Deploying a movie recommender system on Vertex AI

Now, let’s walk through an example of creating a movie recommendation system based on a collaborative filtering type model, deploying it on Vertex AI, and then querying it to get movie recommendations for specific users and movie genre types. The key steps are as follows:

1. Data preparation
2. Model design and training
3. Local model testing
4. Registering the model on Vertex AI
5. Deploying the model
6. Getting predictions

Note

The notebook for this exercise can be found at https://github.com/PacktPublishing/The-Definitive-Guide-to-Google-Vertex-AI/blob/main/Chapter15/Chp-15_Movie_Recommender.ipynb.

Dataset: To train the model, we will use the MovieLens dataset (F. Maxwell Harper and Joseph A. Konstan. 2015. The MovieLens Datasets: History and Context. ACM Transactions on Interactive Intelligent Systems (TiiS) 5, 4: 19:1–19:19. https://doi.org/10.1145/2827872).

The MovieLens dataset is one of...

Summary

In this chapter, we provided a brief overview of recommender systems, different techniques used for building them, and detailed steps for training, deploying, and querying a movie recommender model on Google Cloud’s Vertex AI. Since the key objective was to showcase how you can address a real-world use case using GCP Vertex AI, we kept the core model somewhat simple. But if you are interested in doing a deeper dive into recommender solutions, you can look at courses such as Recommender Systems Specialization on Coursera.

In the next chapter, we will look into another real-world use case around building a vision-based machine learning solution to detect defects during the manufacturing process.

References

To learn more about the topics that were covered in this chapter, take a look at the following resources:

Collaborative Filtering for Movie Recommendations: https://keras.io/examples/structured_data/collaborative_filtering_movielens/

Get started with Vertex AI Model Registry: https://github.com/GoogleCloudPlatform/vertex-ai-samples/blob/main/notebooks/official/model_registry/get_started_with_model_registry.ipynb