A method includes determining signaling sequence data of a user according to signaling plane data of the user, the signaling plane data of the user includes N pieces of signaling data and a timestamp of each of the N pieces of signaling data, and each of the N pieces of signaling data includes a signaling type, determining a first probability value of occurrence of a first network event according to the signaling sequence data of the user and a first supervised prediction model, determining a second probability value of occurrence of the first network event according to expert feature data of the user and a second supervised prediction model, the expert feature data of the user represents a value of at least one preset feature of the user, and determining a predicted occurrence probability of the first network event according to the first probability value and the second probability value.