Apparatus and method for predicting user-item preference based on adaptive elastic network

Abstract

The invention discloses a user-commodity preference prediction device and method based on an adaptive elastic network. The prediction device includes a data preprocessing module, a storage module, a prediction data generation module, and a data output module; it also includes a parameter control module configured to For judging whether the control parameters meet the construction and update conditions of the user-commodity preference prediction model based on the adaptive elastic network; the adaptive parameter control module is configured to construct the prediction model according to the user-commodity rating data and control parameters and Informing the prediction data generation module to perform data training on the prediction model. The prediction device and method provided by the present invention limit the user characteristics through the elastic network, increase the sparsity of the model, better reflect the user's community relationship, and improve the prediction accuracy and accuracy of the model through the efficient restriction of the model through the elastic network. Computationally efficient, it can be widely used in e-commerce platforms that provide personalized services.

Description

technical field [0001] The invention relates to the technical field of computer data processing, in particular to a user-commodity preference prediction device and method based on an adaptive elastic network. Background technique [0002] In the family financial platform, the number of household users is huge and the types of commodities (for example, household financial products) are very rich. It is difficult for users to choose the most suitable financial products for their families among the numerous commodities. Therefore, a huge user-product rating matrix can be formed through the user's rating of the product, and the rating can be used to predict the user's preference for the product. Generally speaking, the user-item rating matrix is ​​an extremely sparse high-dimensional matrix, because there are many types of items, and it is impossible for each user to rate all items one by one. [0003] According to the historical scores on the e-commerce platform, we can unders...

AI Technical Summary

Problems solved by technology

Because there is a large loss of precision when processing some data, it is only based on L 2 Regularization restricts user characteristics, it is difficult to predict user preferences with high precision, and it cannot achieve better service recommendation

That is: (1) During the model construction, during the process of extracting and expressing the characteristics of each user in the latent feature space, since each user has different preferences, only L 2 To limit user characteristics, it is impossible to accurately express the product relationship trend corresponding to each user, so it is necessary to make more scientific and strict restrictions on user characteristics; (2) only use L 2 To limit the objective function, there is a large error in mapping the user's characteristics to the feature space, so that there is a correlation between the user's characteristics and some non-correlated data clusters in the latent feature space, so that the increase in the entire process of modeling many uncertain factors and noise, greatly reducing the predictive performance of the model

Images