An Experimental Study of Recommendation for Wireless Edge Caching

Recommendation-enhanced wireless edge caching has been receiving increasing attention due to its great potential in convincing users to request the cached contents by the means of recommendation. However, since the recommender system is biased towards the cached contents, which may no longer best match the users' preference, doubts about the effectiveness of the caching-oriented recommendation naturally arise. In this paper, as a first attempt, we conduct a campus-wide experimental study to justify some commonly used assumptions about the impact of the caching-oriented recommendation on user behavior, namely sequential requests for recommendation lists, long-term trust in recommendation, importance of video's position in a related list, and view completion rate of recommended videos. The experimental and simulation results verify the assumptions and show the advantages of recommendation-enhanced caching, shedding light on its practical application.