Energy Optimization for 5G Network Slicing by Using Convolutional Neural Network Based Gated Recurrent Unit with Harris Hawks Optimization

In recent years, 5G network slicing enabled customized services for different applications but requires efficient energy management. Optimizing resource allocation reduces power consumption while maintaining performance. Traditional approaches for optimizing energy in 5G network slicing have faced several challenges which include resource allocation complexity and high computational overhead. Therefore, this research proposes Convolutional Neural Network based-Gated Recurrent Unit with Harris Hawks Optimization (CNN-GRU-HHO) to optimize energy consumption in 5G network slicing. The data is collected from Unicauca IP flow Version 2 dataset, which consists of 3,577,296 observations and 87 attributes related to IP flow characteristics. Next, the collected data undergoes preprocessing with mean imputation to handle missing values and one-hot encoding is applied to transform data and enhance feature representation effectiveness. Subsequently, CNN-GRU is employed for network slicing classification and optimized using HHO to improve performance based on adaptive learning rate. The proposed CNN-GRU-HHO achieved superior performance in terms of accuracy (98.21%), precision (96.14%), recall (90.74%) and F1-Score (95.87%) when compared with existing proposes CNN based Long Short-Term Memory (CNN-LSTM).