Workload prediction method and device for service in service cluster

The embodiment of the invention provides a workload prediction method and device for services in a service cluster. The method comprises the following steps: acquiring n + 1 index sequences respectively corresponding to each service in the same historical time period, including n state index sequences corresponding to n system state indexes and load index sequences corresponding to working load indexes; on the basis of the n + 1 index sequences corresponding to all the services, correlation coefficients of the same index between the different services are calculated, similarity graphs corresponding to the same index are constructed, and n + 1 similarity graphs are obtained; nodes in the similarity graph represent services, and connection edges between the nodes correspond to correlation coefficients; processing n + 1 similarity graphs by adopting an attention mechanism to obtain a fused similarity graph; through a first convolutional network, performing graph convolution processing on the initial representati