A method for the middle-term prediction of spot prices for cloud service using multi-layer perceptron is considered in this paper. The analysis of the existing solutions is provided and the model of a standard multi-layer perceptron is presented. The experimental research is fulfilled on the Amazon Elastic Cloud computing spot instances. The obtained prediction results showed a high accuracy of the proposed approach. It allows the users of cloud services to apply the proposed method for the running computational tasks in a most economic way.
cloud services, spot prices, prediction, multi-layer perceptron