Industrial key quality index prediction method based on long-term and short-term dependence of time sequence

The embodiment of the invention provides an industrial key quality index prediction method based on time sequence long and short term dependence, and belongs to the technical field of data processing, and the method specifically comprises the steps: determining a key quality index, and screening out a modeling auxiliary variable; training an auxiliary variable hidden space embedding model; constructing a progressive self-correcting depth condition diffusion model framework; constructing a time sequence long-term and short-term dependence noise prediction model; performing training by taking the noise prediction error as a loss function; noise in data obtained from Gaussian distribution sampling is gradually removed, noise-free target information is obtained, and a median of a key quality index part of the noise-free target information is used as an index prediction value; an auxiliary variable part in the noise-free target information is restored to a sample space, and an obtained data median is used as an au