A Recommender System Based on TensorFlow Framework
Hukam Singh Rana, Thipendra P. Singh · 2023
Over the past few years, personal recommendations have become a part of many online services, such as e-commerce, advertising, and social media. Most businesses estimate a user’s adoption rate based on previous interactions, like purchases and clicks. A recommendation algorithm, collaborative filtering (CF), addresses this issue by assuming that similar users will have similar preferences. Parameterizing users and items implement this assumption to reconstruct historical interactions and estimate user preferences based on the parameters. The embedding of users and items is an essential component of modern recommender systems. Creating deep learning-based recommendations from scratch involves setting up models for experiments that is time-consuming and inefficient. We need to undergo several architectural iterations to determine the optimal set of hyperparameters for a particular use case. TensorFlow recently released a TensorFlow Recommenders (TFRS) package, which makes developing recommender systems far simpler. TFRS is an open-source package based on TensorFlow 2.x that simplifies advanced recommender models’ building, evaluation, and deployment. TFRS allows us to create and assess flexible candidate nomination models and freely include item, user, and context information into recommendation models, etc. This chapter aims to examine TensorFlow recommenders in implementing a recommender system.