Reinforcement learning driven network map region clustering prefetching method

Abstract

The invention provides a reinforcement learning driven network map region clustering prefetching method. Tile small files in the same region are merged into a region clustering large file, a reinforcement learning driven network map region clustering prefetching model is established, the model adds recorded and counted hotspot regions and region space relations into Q learning parameters, and the region corresponding to the maximum Q value direction is a corresponding prefetching region. By caching network map user request regions, the utilization rate of network map server cache space is increased, network map data is cached and prefetched in the form of the region clustering large file, the caching and prefetching quality of tiles is improved, future hotspot regions of the network map are dynamically predicted and cached through the combination of the characteristics of the network map and historical access records, active large-scale efficient prefetching is achieved, dynamic update and accurate prediction are achieved, and the performance of a network map server can be improved.

Description

Technical field [0001] The invention relates to a network map region clustering prefetching method, in particular to a network map region clustering prefetching method driven by reinforcement learning, and belongs to the technical field of network map data prefetching. Background technique [0002] There are many concurrent users of the network map, the large amount of map data, and the long transmission time. Due to the low service quality of traditional network maps, the application of network maps is severely restricted, and a network map server system with a greatly improved performance is required. The prior art network map server still needs to obtain the tile content and data frequently accessed by users from the hard disk every time. This process is expensive, slow, and low in efficiency. The cache of the server is not fully utilized, which seriously affects Improve the efficiency of the server. [0003] If the map data frequently accessed by users is directly cached or p...

AI Technical Summary

Problems solved by technology

There is a cache based on a single tile in the prior art, but the cache unit is small, many tiles need to be cached, the organization and management are relatively complicated, and there is also a lack of prediction of future access hotspot trends on the network map, and there is even a lack of integration with the network map. Access features Early prefetch for future access hotspots

[0006] On the whole, the existing technology mainly has the following defects: First, it lacks the effective utilization of the cache space of the network map server. For the map content and data frequently accessed by users, it still needs to be obtained from the hard disk every time, and the process costs a lot. large, slow, and low efficiency, seriously affecting the work efficiency of the server; second, the network map data are all tile files, and the tile files are small files with a huge number. The organization and management of massive tile data is very complicated, and the Too many small files are not conducive to file management, search and service, and will also lead to too many and complicated cache and prefetch files; third, the lack of advanced prefetch methods for efficient use of server cache, lack of accurate prefetch of network map hotspots, The key method to improve the performance of the network map server is the lack of a method to dynamically predict the future hotspot areas of the network map in real time in combination with the characteristics of the network map itself and historical access records

Images