Comparison of Real-Time and Batch Job Recommendations

Abstract

Collaborative filtering recommendation systems are traditionally trained in a batch manner and are designed to produce personalized recommendations for a large number of users at the same time. However, in many industrial use cases, it is reasonable to produce recommendations in real time, taking account of very recent user interactions. In this work, we present the implementation of batch and real-time recommendation systems using the example of the RP3Beta model, a simple scalable graph-based model that outperforms multiple more advanced models. Our approach can be utilized by any recommendation system if user-to-item recommendations can be obtained based on item-to-item recommendations. We show that it covers multiple common recommendation models, especially collaborative filtering approaches where user features are not available. We also provide the results of A/B tests comparing these two approaches in a real-world scenario of a job recommendation task, conducted with almost 200,000 OLX users. We report at least 10% more users applying for recommended job ads when using a real-time instead of a batch approach. We believe that our results can help other organizations to take informed decisions about whether to make the effort of moving from a batch to a real-time recommendation setting.INDEX TERMS A/B tests, collaborative filtering, job recommendations, real-time recommendations, RP3Beta.

The Impact of Shared Information Presentation Time on Users’ Privacy-Regulation Behavior in the Context of Vertical Privacy: A Moderated Mediation Model

“…It can also refer to the time that recommended content based on cross-platform shared data updated by the recommendation algorithms is presented to the users. In a real-time or streaming recommendation situation [ 7 , 8 , 9 ], the data’s update and presentation time varies depending on the recommendation algorithms, which are typically based on the combination of two strategies that generate the updated recommendation results: (1) regular collection of user data for offline modeling and updating, or (2) online real-time update based on the user’s behavior in the recent seconds [ 7 , 8 , 9 , 39 ]. The latter is quite popular, with several mainstream apps, including Tiktok and Alibaba, adopting real-time recommendation algorithms.…”

“…For example, several mainstream applications (apps), such as Facebook, Twitter, and WeChat, have an increasing number of related parties and authorized partners with whom they share certain low-sensitivity information through the third-party service agreements that users have to authorize before using the service [ 3 , 4 , 5 , 6 ]. Concurrently, real-time recommendations with faster response capabilities [ 7 , 8 , 9 ] are widely adopted in these apps. The combination of platforms’ data stream sharing structure and real-time algorithms has improved the accuracy and responsiveness of apps’ personalized services [ 10 ].…”