Cloud computing provides organizations with pay as you go model and on demand access to computing resources. To optimize the performance of the cloud, resource management should be done with proper allocation techniques, scaling and monitoring. In this paper, we have proposed a novel approach using a Self- normalizing Neural Network(SNN) and Transformer model to predict future resource needs in cloud computing. SNN are used to stabilize the training process whereas transformers are used to capture the long term temporal dependencies. This hybrid model reduces the complexity along with mitigating the problem of vanishing and exploding gradients which makes it ideal for dynamic resource allocation.