Scaled Conjugate Gradient Backpropagation based SLA Violation Prediction in Cloud Computing

Cloud Services are highly scalable and dynamic. Providing high availability and low latency with high data rate is a big challenging task for service providers. Any deviation in the Service Level Agreement (SLA) between service providers and customers is service level Violation (SLV). In the case of SLV, the cloud service provider has to pay the penalty to the customers in terms of money or additional services as per the SLA. Hence SLA prediction is of paramount importance for both customers and service providers. In this paper, we have used a Scaled conjugate gradient neural network to make predictions on SLA Violations. We have used real-world data which seems to be relevant to the real-world scenarios. Oversampling technique is used to overcome the problem of the biased dataset. Results obtained from the experiments have undergone 3-fold validation. Classification accuracy of SCG algorithm is found to be 96.76% and 93.23% for two datasets which may be considered as an efficient prediction model.