A matrix decomposition method and device based on user project scoring

Abstract

The embodiment of the invention provides a matrix decomposition method and device based on user project scoring. The method comprises the following steps: constructing a mean square error loss function according to a to-be-decomposed user project scoring matrix and a decomposition relation between a user factor matrix and a project factor matrix; Iteratively updating the mean square error loss function for multiple times until the iteratively updated mean square error loss function meets a preset iteration condition, and the iteration step length corresponding to each variable element in different iteration processes is dynamically changed along with the difference of the iteration times; And determining the value of each updated variable element in the last iteration as a decomposition optimization value of each variable element in the user factor matrix and the project factor matrix. The device is used for executing the method. According to the method and the device provided by the embodiment of the invention, while the decomposition speed is improved, the accuracy of the decomposed user factor matrix and project factor matrix is improved.

Description

technical field [0001] The embodiments of the present invention relate to the field of data processing, in particular to a matrix decomposition method and device based on user item ratings. Background technique [0002] The hidden factor matrix decomposition algorithm is to decompose a user-item matrix (usually a high-dimensional sparse matrix, that is, many elements in the matrix are missing and unknown) into two matrices, a user factor matrix and an item factor matrix, and the decomposed two A matrix is ​​randomly assigned at the beginning, and then the loss function is constructed according to the value of the user-item matrix (ie, the actual value) and the value of the product of the corresponding decomposed two matrices (ie, the predicted value). In order to make the predicted value as close as possible to the actual value, it is necessary to make the loss function as small as possible. [0003] At present, the stochastic gradient descent method can be used to find the...

AI Technical Summary

Problems solved by technology

[0004]However, in the existing hidden factor matrix factorization algorithm, since the variable elements in the user factor matrix and item factor matrix follow the same direction of the negative gradient of the loss function The movement of the learning rate (step size) will lead to repeated iterations to approach the minimum value, and the decomposition speed is slow; or cause the minimum value point to be missed, and accurate decomposition results cannot be obtained

Images