Collaborative filtering to improve navigation of large radiology knowledge resources

Collaborative filtering is a knowledge-discovery technique that can help guide readers to items of potential interest based on the experience of prior users. This study sought to determine the impact of collaborative filtering on navigation of a large, Web-based radiology knowledge resource.

Collaborative filtering was applied to a collection of 1,168 radiology hypertext documents available via the Internet. An item-based collaborative filtering algorithm identified each document's six most closely related documents based on 248,304 page views in an 18-day period. Documents were amended to include links to their related documents, and use was analyzed over the next 5 days.

The mean number of documents viewed per visit increased from 1.57 to 1.74 (P < 0.0001).

Collaborative filtering can increase a radiology information resource's utilization and can improve its usefulness and ease of navigation. The technique holds promise for improving navigation of large Internet-based radiology knowledge resources.