A Machine Learning-Based Small Station Caching Method for Ultra-Dense Networks

Abstract

The invention discloses a machine learning-based ultra-dense network small station caching method. First, the K-means clustering method is introduced to analyze historical access data during off-peak access periods, and the space-time pattern of file requests is mined. Preference clustering, find out popular files in the small site, realize personalized cache between classes and predictive cache within classes, and use historical access data and clustering results to construct a training set for new file classification; and then at the peak During the access period, the k-nearest neighbor classification method is introduced to periodically classify the emerging new files and cache them in the small station category that prefers such files; finally, combine the historical popular files in various small stations and the emerging new files to formulate real-time updates caching strategy. The invention formulates a cache strategy based on machine learning, can make full use of the limited cache space of the small station to store the files most needed by the people served by the small station, significantly reduce the load of the system backhaul link, and greatly improve user satisfaction.

Description

technical field [0001] The invention belongs to the technical field of network communication, and relates to a caching method of a base station, and more specifically, to a caching method of a small station in an ultra-dense network based on machine learning in a wireless communication system. Background technique [0002] The rapid increase of mobile traffic in the 5G (the fifth generation) network has brought great challenges to mobile network operators. As one of the candidate technologies for 5G, the ultra-dense network technology of densely deploying small cells at the same frequency within the coverage area of ​​macro cells can effectively improve spectrum efficiency and system throughput. Small and medium-sized stations in ultra-dense networks are often deployed in some difficult-to-reach locations, which brings difficulties to the installation of optical fiber backhaul links connecting small stations and core networks. To solve this problem, wireless backhaul technol...

AI Technical Summary

Problems solved by technology

[0002] The mobile communication traffic in the 5G (the fifth generation) network is soaring, which brings great challenges to mobile network operators

[0003] Most of the existing caching technologies are based on traditional optimization algorithms to formulate caching strategies, and these works are often based on strong assumptions, making caching strategies difficult to apply to actual systems

Moreover, these caching strategies are generally formulated based on historical access data. Considering that new files will be accessed in large numbers during the peak access period in the network, caching strategies are only formulated based on the patterns obtained from historical access data, which cannot effectively utilize the limited cache. space

Images