Çok Büyük Konumsal Verinin Görselleştirilmesinde Karmaşıklık Yönetimi

Abstract

Data visualization is becoming more challenging by the day due to a continuous increase in the size of data to be visualized. Geographical data is no exception, especially considering that smart phones enable almost anybody to produce location data. However, visualization of such large data has to be done on a map, which be- comes crowded, cluttered and unreadable at lower zoom levels. If we increase the zoom level, we can display more details, but we lose the completeness of visuals. In this study, we consider a binary feature to be visualized on a map. We apply three different approaches to recognize feature clusters within the data. Each cluster then corresponds to a geographical region and one of the two feature values. The visualization done like this results in a minor amount of information loss. We compare these three methods with respect to entropy gain, memory and speed by measuring this loss in terms of information entropy. Also, we provide detailed enumerative results under different visualization scenarios.