Click rate prediction method, system and device and readable storage medium

The embodiment of the invention discloses a click rate prediction method, system and device and a readable storage medium, and the method comprises the steps: receiving a commodity click rate prediction request which carries user data and commodity data; converting the user data and the commodity data into a feature vector set, and performing coding conversion on the feature vector set to obtain a coding feature set; inputting the coded feature set into a click rate prediction neural network to calculate a comprehensive attention score between each pair of coded features, the comprehensive attention score being obtained according to a pairwise interactive attention score and a feature domain attention score of each pair of coded features; obtaining an output value of the attention layer in the residual network according to the comprehensive attention score and the feature embedding representation corresponding to the comprehensive attention score; and inputting the output value into a full connection layer to