Neural network recommendation method based on multi-dimensional cloud and user dynamic interest

Abstract

The present invention relates to the fields of information retrieval and data mining, and in particular to a neural network recommendation method based on multi-dimensional cloud and user dynamic interests, including: obtaining user ratings for items and user rating time and performing preprocessing; using preprocessed data Build a user-based multidimensional cloud model and an item-based multidimensional cloud model to obtain user-based predictive scores and item-based predictive scores; use user-based predictive scores and item-based predictive scores as the trained neural network score prediction model Input data to obtain the final prediction score; the invention not only effectively utilizes user data, but also improves the disadvantages of the personalized recommendation method in the data sparse scenario.

Description

technical field [0001] The invention relates to the fields of information retrieval and data mining, in particular to a neural network recommendation method based on multi-dimensional cloud and dynamic interests of users. Background technique [0002] The continuous development of Internet and Web2.0 technology has generated a large amount of information data for users, which meets the needs of users for information in the information age. However, information overload has a serious impact on information consumers and information producers. The most effective solution to information overload The best method is the recommendation system. [0003] As an intelligent information service method, recommender systems are widely used in e-commerce, social networks, movies and videos, music, personalized emails and advertisements. The recommendation system is represented by the personalization technology of electronic services. It analyzes and calculates the historical behavior data...

AI Technical Summary

Problems solved by technology

[0004] Although the recommendation technology based on collaborative filtering has a good recommendation effect, in recent years we have observed that the availability of digital information, resources and online content has exploded, and users' evaluation information on items is missing a lot, resulting in data sparsity. Therefore, the traditional collaborative filtering recommendation method is no longer suitable for the current scale of the network

And due to the individual differences of users, the randomness and fuzziness of user evaluation information are caused. The randomness of user rating information means that users change their rating behavior under specific rating behaviors. The fuzziness of user rating information means that users can only use 1-5 The evaluation of the item is described in terms of points, which is vague and inaccurate. It is not more accurate to express the concept of qualitative language, which makes it difficult to predict the score with low accuracy.

Images