Optimizing Cloud Load Forecasting with a CNN-BiLSTM Hybrid Model

Abstract

Cloud computing has emerged as a cornerstone for modern industries, offering scalable and flexible resources to meet growing computational demands. However, managing fluctuating workloads in cloud data centers poses significant challenges, often leading to inefficient resource allocation and energy wastage. This paper proposes a novel hybrid model combining Convolutional Neural Networks (CNN) and Bidirectional Long Short-Term Memory (BiLSTM) networks to address the problem of cloud load prediction. The CNN-BiLSTM model leverages the strength of CNNs for spatial feature extraction and BiLSTMs for capturing temporal dependencies in cloud workload data, providing improved prediction accuracy over traditional models. A comprehensive comparison of the CNN-BiLSTM model against other deep learning architectures, including Backpropagation (BP), LSTM, and CNN-LSTM, demonstrates significant enhancements in prediction performance. The model's ability to predict cloud load more accurately can contribute to more efficient resource management in cloud environments.