A model-based data processing method and apparatus, an electronic device, a computer-readable storage medium, and a computer program product. The method comprises: extracting a first training sample set to obtain a second training sample set and a query video, wherein the first training sample set comprises different types of video samples; processing the second training sample set by means of an embedded layer network in an action recognition model to obtain a first frame feature sequence; processing the query video by means of the embedded layer network to obtain a second frame feature sequence; processing the first frame feature sequence by means of a timing relationship network in the action recognition model to obtain a first timing relationship descriptor; processing the second frame feature sequence by means of the timing relationship network to obtain a second timing relationship descriptor; and adjusting model parameters of the action recognition model according to the first timing relationship descriptor and the second timing relationship descriptor.