Friend recommendation (FR) becomes a valuable service in location-based social networks. Its essential purpose is to meet social demand and demand on obtaining information. The most of current existing friend recommendation methods mainly focus on the preference similarity and common friends between users for improving the recommendation quality. The similar users are likely to have similar preferences of point-of-interests (POIs), the kinds of information they provided are limited and redundant, can not cover all of the target user’s preferences of POIs. This paper aims to improve amount of information on users’ preferences through FR. We give a definition of friend recommendation considering preference coverage problem (FRPCP), and it is also one NP-hard problem. This paper proposes the greedy algorithm to solve the problem. Compared to the existing typical recommendation approaches, the large-scale LBSN datasets validate recommendation quality and significant increase in the degree to preferences coverage.