Estimating Category of Pois Using Contextual Information

Abstract

With the popularity of smart phones with a GPS function, location-based applications are widely used. The location-based applications require specific information of POIs (Points of Interest) such as name, category, and exact location as their basic information. However, it costs high to gather such information manually. Furthermore, POIs are located in large geographic areas. For these reasons, the information of POIs should be automatically gathered. In this paper, we propose a method to estimate the category of POIs automatically using two kinds of POI contexts. The two contexts are internal and external information of POIs. The category of POIs is sometimes exposed by their names. Thus, their name itself is used as internal context in estimating POI category. When the category can be determined by the names, the documents that describe POIs are used as external context. Such documents are widely available through internet review sites and contain various kinds of information for POIs such as location, service satisfaction, and menu. Thus, the category of POIs can be also estimated by analyzing this information. We train a machine learning algorithm, support vector machine for each kind of information, and then combine both SVMs. According to the experimental results, the proposed method shows accuracy of 70.35% for 20 POI categories.Keywords: Classification, POI Category Estimation, Point of Interest