Mobile Systems Design Lab Principal Investigator:
Professor Sujit Dey

University of California, San Diego

Overview Projects People Publications News Contact Us Links

Building Scalable Recommendation Systems for Enterprise Knowledge Workers

 

Overview

Enterprise knowledge workers have been overwhelmed by the growing rate of incoming data in recent years. In this paper, we present a recommendation system with the goal of helping knowledge workers in discovering useful new content. Specifically, our system builds personalized user models based on file activities on enterprise network file servers. Our models use novel features that are derived from file metadata and user collaboration. Through extensive evaluation on real world enterprise data, we demonstrate the effectiveness of our system with high precision and recall values. Unfortunately, our experiments reveal that per-user models are unable to handle heavy workloads. To address this limitation, we propose a novel optimization technique, Active Feature-based Model Selection, that predicts the user models that should be applied on each test file. Such a technique can reduce the classification time per file by as much as 23 times without sacrificing accuracy. We also show how this technique can be extended to improve the scalability exponentially at marginal cost of prediction accuracy, e.g., we can gain 169 times faster performance on average across all shares by sacrificing 4% of F-score.

Figure1: Overview of our approach

Figure 1 gives an overview of the proposed approach, consisting of an offline step (executed once), and an online step, which is executed for each new file created in the repository. In the Offline step, access patterns of individuals are captures in the form of trained models. In the Online step, as new files are created, the models of all the users in the repository are applied on the file to generate predictions on whether a user is likely to access the file or not.



Problem addressed

How can we build a file recommendation system specifically for newly generated files that is effective in alleviating information overload experienced by enterprise knowledge workers? How can we make the system scale to large number of users and to large file generation rates?



Publications
 
Below are the publications based on the above work:

  1. C. Verma, S. Bhatkar, M. Hart, A. Parker-Wood, S. Dey, "Access prediction for knowledge workers in enterprise data repositories". in Proc. of International Conference on Enterprise Information Systems (ICEIS), Barcelona, Spain, Apr. 2015. Nominated for best student paper award PDF

  2. C. Verma, S. Bhatkar, M. Hart, A. Parker-Wood, S. Dey, "Improving Scalability of Personalized Recommendation Systems for Enterprise Knowledge Workers". Accepted for IEEE Access 2015. . Url will be available soon.
See other related papers here: Here





People
 

Chetan Verma
Graduate student


Sujit Dey
Adviser

Michael Hart (Symantec Research Labs)
Collaborator

Sandeep Bhatkar (Symantec Research Labs)
Collaborator

Aleatha Parker-Wood (Symantec Research Labs)
Collaborator




back to top






Sponsors

Symantec Research Labs, Mountain View, CA