The invention belongs to the technical field of convexity prediction, and in particular relates to a board convexity prediction method based on kernel partial least squares combined with a support vector machine, comprising the following steps: S1, using a high-precision monitoring device to collect on-site data; S2, collecting the collected Data preprocessing; S3, establishment of KPLS regression prediction model; S4, establishment of KPLS-SVM plate convexity prediction model. In the present invention, the data-driven algorithm is used as a mathematical tool, and abnormal values can be removed from a large amount of collected on-site rolling process data. The present invention establishes a continuous strip rolling strip convexity prediction model based on the kernel-partial least squares method combined with the support vector machine, The prediction of the crown of the strip continuous rolling is realized, and the invention adopts the particle swarm optimization algorithm to optimize the built model, and further improves the prediction accuracy of the strip continuous rolling crown. The invention is used for the prediction of the crown of the continuous rolling strip.