Understanding Users' Preference on Big Data

There are a lot of records in the era of big data, but it is difficult for users to find their favorite record. There exist many different ways of helping a user to solve this problem. One is an interactive approach which asks the user a number of questions (e.g., presenting two records and asking the user to choose a more preferable record) and then learns the user's preference automatically. Another is a log-oriented approach which analyzes the user's log data recorded by the system and learns the user's preferences via some data mining and deep learning approaches. After a user's preference is found, his/her favorite record can be easily determined.

The following figure shows a scenario of our interactive approach. Consider that a user would like to find his/her favorite car. After asking the user about 5-7 questions, our system is able to know the user's preference and find his/her favorite car. In contrast, state-of-the-art methods in the literature of machine learning and database require asking the user about four times as many questions as in our system, which is tedious for the user.

The figure below shows the user interface of our demo system that helps a user find his/her favorite car.

Consider the log-oriented approach. Whenever a user visits a website, all activities of the user can be tracked. We proposed to use a recurrent neural network-based approach to model a user's preference in order to find the record that s/he is interested in. Our experimental result shows that the accuracy of our proposed approach is better than the best-known result.

Prof. Raymond Chi-Wing Wong has led a research team working on this project. Details about this project can be found in the following publications:

1. Tianwen Chen and Raymond Chi-Wing Wong, "Session-based Recommendation with Local Invariance", the 19th IEEE International Conference on Data Mining (ICDM 2019), Beijing, China on November 8-11, 2019
2. Min Xie, Raymond Chi-Wing Wong, and Ashwin Lall, "Strongly Truthful Interactive Regret Minimization", the 2019 ACM Conference on Management of Data (SIGMOD), Amsterdam, The Netherlands on 30 June-5 July, 2019
3. Min Xie, Tianwen Chen and Raymond Chi-Wing Wong, "FindYourFavorite: An Interactive System for Finding the User's Favorite Tuple in the Database", the 2019 ACM Conference on Management of Data (SIGMOD), Amsterdam, The Netherlands on 30 June-5 July, 2019 (Demonstration Paper)