1 Collaborative Filtering - A Group Profiling Algorithm for Personalisation in a Spatial Recommender System Ali Tahir, Gavin McArdle, Andrea Ballatore, Michela Bertolotto School of Computer Science and Informatics University College Dublin Belfield, Dublin 4, Ireland ali.tahir@ucd.ie , gavin.mcardle@ucd.ie , andrea.ballatore@ucd.ie , michela.bertolotto@ucd.ie Abstract. As the quantity of geospatial information rapidly increases, information overload in the spatial domain is becoming a serious issue. Often the amount of information being displayed on digital maps makes it difficult to determine useful content. In order to assist in resolving this problem, personalisation techniques have been developed. The most effective techniques implicitly monitor user interactions with map interfaces and content in order to infer user interests. This permits the map to be adapted to suit individual users by highlighting or displaying a subset of available content. The work presented in this paper builds on this paradigm to refine map adaptations by using group profiling techniques. Such techniques identify similar users to the target user and utilise their interests as an indicator of possible interests for the target user. The map can then be adapted accordingly. The approach has been effective in the non-spatial domain however it has not been widely studied in the spatial context. The methodology behind this technique is presented in this paper while the approach is demonstrated through a case study of a map navigational assistant. 1 INTRODUCTION Spatial information overload has emerged as a new issue in the spatial context in recent years. Extracting meaningful information from large spatial datasets and portraying this on a map results in an efficient way to recommend map items to users. Such a recommender system can be built by monitoring user interactions with the system in order to generate user profiles. These profiles can then be used as a basis on which to recommend map content to users. A system has already been implemented to operate with individual user profiles and recommend items of spatial interest to the user (Ballatore et al. 2010). This paper presents an extension of this work and the development of a prototype which recommends spatial items based on group profiles using collaborative filtering techniques. The paper is structured as follows: section 2 outlines related work in the area of user profiles and map personalisation. Section 3 discusses group profiling approaches with their relative strengths and weaknesses. Section 4 elaborates on the methodology and implementation of the group profiling algorithm. In section 5, conclusions and future work are discussed.