Enhancing Heterogeneous Data Processing Efficiency in 5G Networks using Transfer Learning

The heterogeneous nature of data formats, sources and network requirements represents a significant challenge for the heterogeneous data processing that is a hallmark of the rapidly developing 5G network. The efficient processing of this data is of critical importance in maintaining low latency and high-quality real-time services. This study proposes the implementation of a transfer learning-based model with the objective of enhancing the processing efficiency of heterogeneous data in 5G networks. The model transfers the learned features to a new diverse dataset through pre-training on similar data types, thereby optimising the data processing process and avoiding the necessity for a large number of retrainings. The model comprises two key components: adaptive feature mapping and a flexible architecture that enables adaptation to new data features while minimising computational costs. The results of the experimental analysis demonstrate that the model has a considerable impact on reducing processing time and enhancing data throughput.