Improving recommendation by connecting user behavior in temporal and topological dimensions

The rapid development of Internet provides many high quality data sets for understanding human online behavior patterns. Many analysis tools from complexity science such as complex networks and human dynamics have been applied to study human online activities. However, in most existing works, the to...

Full description

Saved in:
Bibliographic Details
Published in:Physica A Vol. 585; p. 126378
Main Authors: Li, Heyang, Zeng, An
Format: Journal Article
Language:English
Published: Elsevier B.V 01.01.2022
Subjects:
Behavioral dynamics
Complex networks
Power-law distribution
Recommendation algorithms
Power-law distribution
Recommendation algorithms
Complex networks
Behavioral dynamics
ISSN:0378-4371, 1873-2119
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The rapid development of Internet provides many high quality data sets for understanding human online behavior patterns. Many analysis tools from complexity science such as complex networks and human dynamics have been applied to study human online activities. However, in most existing works, the topological and temporal features of human online activities are analyzed independently. In this paper, we connect these two dimensions by investigating the relations between online users inter-event time and the network distance between the items selected by users. We find that users will choose items with longer network distance when the inter-event time is long, and vice versa. This finding is then applied to improve the recommendation process where recommendation diversity should be given more weight when users return from a long inactive period. Finally, we use a trade-off between accuracy and complexity to further improve the recommendation algorithms.
ISSN:0378-4371
1873-2119
DOI:10.1016/j.physa.2021.126378