Click rate prediction model training method, commodity recommendation method and device

The invention provides a click rate prediction model training method, a commodity recommendation method and a commodity recommendation device, and relates to the technical field of commodity recommendation. According to the scheme, a training sample set corresponding to sample users is obtained, and each training sample in the training sample set comprises sample commodity features, sample user features, full-platform commodity interaction behavior sample sequence features and aggregation page commodity interaction behavior sample sequence features; performing model training based on the training sample set to obtain a generated target click rate estimation model; the target click rate estimation model is used for outputting first click rate estimation data and second click rate estimation data, and the first click rate estimation data is used for generating a first commodity recommendation list corresponding to the first display page; the second click rate estimation data is used for generating an aggregatio