Score distribution prediction method and device

Abstract

The invention provides a score distribution prediction method and device. The method comprises the following steps: constructing an objective function comprising a symbol function for describing the correlation between two score potential factor vectors and a distance function for calculating the distance between the two score potential factor vectors; inputting training set data, determining a scoring potential factor matrix during current iteration, and calculating a scoring potential factor matrix during next iteration according to the scoring potential factor matrix to obtain a new matrix;substituting the new matrix into an objective function, and solving a first derivative of the objective function; and when the value of the F norm of the first derivative at the new matrix is smallerthan a preset termination threshold value, obtaining a score distribution prediction result of the to-be-predicted sample according to the new matrix, otherwise, taking the new matrix as a score potential factor matrix during current iteration again, and repeating the steps. According to the scheme, the symbol function and the distance function are introduced into the objective function, the correlation between the scoring labels is considered, and therefore the prediction accuracy can be improved when the scoring distribution is predicted based on the correlation.

Description

technical field [0001] The invention relates to the field of computer technology, in particular to a score distribution prediction method and device. Background technique [0002] Usually, the score distribution of the samples to be released can be predicted through score distribution prediction. Therefore, based on certain foresight, the investment risk can be reduced as much as possible. Take the film industry as an example. As a global industry worth tens of billions of dollars, thousands of new films are launched into the market every year, but not all films are successful at the box office. Therefore, both filmmakers and movie theaters need to predict how much audiences will like a new movie before it is released. [0003] At present, the rating distribution of movies can be used to represent the audience's liking for a certain movie. Naturally, the LDL (Label Distrubition Learning, label distribution learning) method can be applied to this field. LDL is a generaliz...

AI Technical Summary

Problems solved by technology

It can be seen that the prediction accuracy of the existing implementation is not high

Images