A Novel Regression System for Predicting LTE Network Performance Metrics

Abstract

The invention discloses a novel regression system for predicting LTE network performance indexes, comprising a cluster module, a regression module, a selection module and a prediction module. The novel regression system operates as the following steps of performing clustering on cells to obtain k clusters , executing a plurality of various regression algorithms on each collection group, choosing the algorithm having a smallest error rate of each cluster as the optimal regression algorithm , obtaining an optimal number k of clusters which have low error values and high separation degrees through combination of the error rate ER and the collection group separation degree aggregation value Sep, and utilizing the obtained optimal collection group number k and the optimal regression algorithm of each collection group to perform prediction on the LTE KPI consumed by the network resource of the small community. The invention not only can obtain the network resource consumption state of each network cell in the mobile communication under the premise of considering the difference of each community, but also can predict the future trend of the LTE KPI index through choosing the regression algorithm.

Description

technical field [0001] The invention relates to a novel regression system for predicting LTE network performance index. Background technique [0002] With the rapid development of LTE networks, the network data created by people has increased exponentially. Data-based consumption services have also become more diverse, such as web browsing, video communication or streaming media, and the popularization of smart terminals. However, the accessibility of the network is further deteriorating due to infrastructure deployment limited by network capacity and excessive consumption of network resources by people. Therefore, mobile operators must reasonably plan network capacity. Considering the high investment cost of network infrastructure and the limitless expansion of network capacity, in order to avoid network overload, local deployment of network capacity must be controllable. Therefore, it has become the most economical and effective method to use big data prediction method ...

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to provide a new type of regression system for predicting LTE network performance indicators to solve the problem in the prior art that when all the cell data are combined, the differences between the cells are lost. After all the cells are gathered together Averaging the data will lead to problems such as lack of information in the measurement data

Images