A switching algorithm based on machine learning in a UDN

Abstract

The invention discloses a switching algorithm based on machine learning in a UDN, and the algorithm comprises the following steps of A when a mobile device needs to be switched to a new micro cell each time, firstly reporting a feature information report of the position where the mobile device is located to a macro base station, and discretizing the feature information by the macro base station; Brespectively training the discretized feature information data by adopting four learners of machine learning decision tree prediction, neural network prediction, SVM prediction and random forest prediction to obtain each training model; C making a decision on each prediction result by using a majority voting method to obtain a decision result; and D when the decision result is that the mobile device needs to be switched and a pre-switching condition is met, using the macro base station to send a pre-switching request to a pre-switched target micro cell, and using the target micro cell to start to prepare resources for the mobile device and implement switching. According to the present invention, the unnecessary switching is reduced, and the average time delay of the system in the ultra-dense network is reduced.

Description

technical field [0001] The invention belongs to the technical field of communication networks, and in particular relates to a switching algorithm based on machine learning in UDN. Background technique [0002] In order to cope with the thousands-fold increase in the data volume of the next-generation wireless communication network, Ultra Dense Network (UDN) is considered to be one of the most promising key technologies and an inevitable trend of future mobile communication. In the UDN, the mobile device constantly switches between cells, resulting in a relatively large switching delay, which brings a large burden to the system. The prediction of user switching can improve the performance of switching. However, for the traditional prediction method based on linear programming, the prediction accuracy is not enough, especially in the face of increasingly dense and complex networks. Not enough to make a positive impact on the performance improvement of switching. As for the p...

AI Technical Summary

Problems solved by technology

[0005] In addition, although a simple machine learning algorithm such as SVM can predict the mobility of terminals relatively accurately in real time, there are often many features that affect user mobility. In the case of multiple features, SVM often cannot obtain a better performance

In the UDN network, facing dense base stations and frequent handovers often require more complex user information characteristics, and the handover prediction made by a single algorithm cannot accurately

Images