Cross-regional communication flow prediction method based on heterogeneous model reuse

Abstract

The invention discloses a cross-region communication flow prediction method based on heterogeneous model reuse. The method comprises a data acquisition step, a model reuse step, a new region model training step and a new region model test step. The method comprises the following steps: firstly, collecting communication data of different regions, including feature data and meta-information description data of features; training a model on the previous regional data, and mapping the heterogeneous model into the feature space of the current region by adopting a model reuse and optimal transportation technology; and finally, establishing a new model on the new area by using a small amount of communication data, soso to train a usable model with good performance by using a small amount of marked data. According to the method, the difficulties that feature spaces of new and old regions are different and only data in the new region is difficult to establish a new model can be solved, and themethod occupies few resources and is high in applicability in the implementation process.

Description

technical field [0001] The invention relates to a traffic forecasting method in the case of changes in communication characteristics of different areas in an open dynamic environment, in particular to a cross-area communication traffic forecasting method based on heterogeneous model reuse. Background technique [0002] In the field of communication, how to quickly and accurately predict the communication traffic of an area in the next period of time is very important. For example, when the communication traffic in an area is very low in the next few hours, the services of some communication base stations can be turned off, so as to reduce the Energy consumption, cost reduction and other purposes. However, the real environment is mostly open and dynamic, and there will be great differences in base station facilities, services provided, and communication quality in different regions, which leads to inconsistencies in the features that can be obtained in different regions. The...

AI Technical Summary

Problems solved by technology

Previous model reuse technologies are mostly limited to application scenarios where old and new regions have the same characteristics, and are not suitable for open and dynamic environments where characteristics change, especially for application scenarios such as cross-region traffic forecasting

Images