In order to explore the action recognition, tracking, and optimization analysis of the training process based on the SVR model and multimedia technology, the author proposes based on the radial basis function model, researching a new surrogate model technology-support vector regression (SVR). We first introduce the basic principles of SVR, select the parameters of SVR, and then elaborate the basic steps of SVR modeling. Then, we design and optimize application examples through numerical example multimedia technology; the validity of the support vector regression method is verified. Experimental results: the comparison of SVR1 and SVR2 shows that the utilization of multiscale timing feature maps should occur after tem (SVR2) rather than being directly fused in the feature dimension (SVR1), mainly because small-scale information affects the resolution of large-scale information; on data sets such as ActivityNet, in order to verify the effectiveness of SVR and DR-Dvc algorithms, the performance of the proposed algorithm and the baseline before improvement and the current mainstream algorithm are respectively compared. Experimental results show the proposed algorithm has a significant performance improvement compared to before the improvement; at the same time, it is better than most current mainstream algorithms, which proves the feasibility and effectiveness of the algorithm. Describing the introduction of regression can effectively improve the performance of sequential action proposals and event description algorithms, and compared with the current mainstream methods, it has certain performance advantages.