Topic Title:

Exploring Users’ Interests to Recommend Novel Items

 

Technical Area:

Recommender Systems, User Experience, e-Commerce, Economics, Psychology

 

Background

The technology of recommender systems has significantly improved the efficiency of internet traffic in the past ten years especially in the mobile internet age. The e-commerce landscape has been greatly reshaped with better user experience and the extent of operation automation based on big data. However, efficient recommendation algorithms learn too much from user behaviors especially the recent user behaviors to optimize the traffic efficiency, e.g. CTR, GMV. Therefore the users’ interests are over exploited to recommend users too familiar items. On one hand, it makes users bored and tired to view more items; on the other hand it reduces users’ potential demand without giving users extensible recommendations.

An cutting-edge research is needed to unravel the complicated problem of this topic, no matter from the perspectives of economics, psychology and machine learning.

 

Target

  1. Find a definition of a consistent system of values for recommender systems sufficient to describe the objectives of recommendation especially for a long-term and ubiquitous running of recommender systems.
  2. Successful online experiments to validate this definition.

 

Related Research Topics

  1. User intention models.
  2. Pareto principle.
  3. User experience modeling.
  4. Ecology of E-Commerce platforms.