A Fuzzy logic based ensemble adaptive tile prefetching

Uluat, Mehmet Fatih
Prefetching is a process in which necessary portion of data is predicted and loaded into memory beforehand. The increasing usage of geographic data in different types of applications motivated the development of different prefetching techniques. These techniques are usually developed for specific type of applications such as 2D geographic information systems or 3D visualization applications and crafted for corresponding navigation patterns. However, as boundary between these application types blurs, these techniques become insufficient for hybrid application types such as digital moving maps. This type of applications possess capabilities from both of these domains and exhibit various navigation patterns. Therefore, a group of prefetching techniques should be used together to handle different requirements and navigation patterns. In this study, a priority based tile prefetching approach is proposed which enables ensemble usage of different prefetching techniques at the same time. The proposed approach manages these techniques dynamically through a fuzzy logic based inference engine to increase prefetching performance and to adapt to various behaviors exhibited. This engine performs adaptive decisions about contribution of each technique according to their individual performance and activity level. The results obtained from experiments showed that up to 25% increase in prefetching performance is achieved with proposed adaptive ensemble usage over single technique usage. A generic model for prefetching techniques is also developed and used to describe given approach. Finally, a cross-platform software framework with five different prefetching techniques are developed to let other users utilize proposed approach.
Citation Formats
M. F. Uluat, “A Fuzzy logic based ensemble adaptive tile prefetching,” Ph.D. - Doctoral Program, Middle East Technical University, 2014.