The application provides a computing
power performance bottleneck intelligent optimization method and
system for
smart office application, and the method comprises the following steps: collecting user behavior and
system resource state data through a sliding window, and constructing a structured task context sequence; a resource-aware attention mechanism is used to identify the current office task type and scheduling priority; based on the task type and priority, nodes are screened from a resource prototype
library and a weighted resource path graph is constructed according to a task conditional prior, and a graph-level representation is obtained through graph neural network coding; a structure attention mechanism is used to generate a path-oriented scheduling plan, and a complexity regularization term is introduced to suppress high-cost actions; the task priority is used to dynamically control the instruction execution
rhythm, and the state is resampled and the weight is calculated to quantify the improvement effect of the scheduling on the
bottleneck path. The application is task-centered, realizes closed-
loop optimization from identification to execution, and significantly improves the response fluency of office tasks.