Genetic algorithms for map generalization

Abstract

A map is usually thought of as a tool (describing a part of the world) for the user, but when a map is poorly designed instead of a tool it becomes an obstacle for the user [1]. A common error when designing maps is the amount of information shown. In a map, inefficiency in information communication is proportional to the amount of irrelevant information according to a user query [2]. When a user asks for specific data in the spatial information context, it is common to find information overload [3], that is, the user gets a very difficult to read map, with much useless data; a map content must be appropriate for the map scale and purpose. A user also has to look at a map time enough for him to discern the information within the map [4]. This observation time varies depending on the user’s familiarity and experience with maps, knowledge on the map’s theme and its ability to discern visual signs on the map [1]. So, it is assumed that the better the map design is, the easier it’ll be to read it. Thereby, the motivation for this work is to create personalized maps containing familiar information (well-known places) to the user in order to improve the map readability. The generalized maps (less detailed maps) will be derived from highly detailed maps, thus leading to a spatial generalization problem. As study case, we will work with Points Of Interest (POI) from Mexico City