Deep neural network Top-N recommendation algorithm based on heterogeneous modeling

Abstract

The invention relates to the technical field of deep learning and intelligent recommendation, in particular to a deep neural network Top-N recommendation algorithm based on heterogeneous modeling, and the algorithm comprises the steps: obtaining and constructing data, and obtaining the interaction information of a user and an article; generating an enhanced display and implicit feedback matrix from a heterogeneous information network for original recommendation, namely a user article scoring network, through a construction meta-path; performing information analysis by the MFDNN model, namely inputting the obtained display feedback information and implicit feedback information into the MFDNN model for information analysis to predict a missing user-project interaction score value; embedding and inputting a user item vector obtained by final training of the MFDNN model into the MFDNN model to obtain scores, and performing sorting output; and comparing the score of the article by the user, which is finally predicted by the MFDNN model, with a real user score to obtain the effect of the model. The method can effectively and accurately solve the problem that a user looks for favorite information in massive information.

Description

technical field [0001] The invention relates to the technical field of deep learning and intelligent recommendation, in particular to a Top-N recommendation algorithm of deep neural network based on heterogeneous modeling. Background technique [0002] The term "big data" has become widely known, and the world has gradually entered the "big data era". With the rapid increase in the amount of information, many users turn to recommendation algorithms when looking for information on learning resources, movies, music, popular events, and other fields. Although the data generated by various Internet applications contains rich information, ineffective management of these data can lead to problems such as information overload. [0003] In order to deal with massive information in the network, modeling technology has great potential in quickly and accurately finding the most popular information; however, the existing problems: 1. A large amount of heterogeneous information network ...

AI Technical Summary

Problems solved by technology

Although the data generated by various Internet applications contains a wealth of information, ineffective management of this data can lead to problems such as information overload

[0003] In order to deal with massive information in the network, modeling technology has great potential in quickly and accurately finding the most popular information; however, the existing problems: 1. A large amount of heterogeneous information network data hides the comprehensive and detailed information of items information

Therefore, mining and analyzing valuable information in heterogeneous information networks is a key challenge; 2. The rapid expansion of heterogeneous information networks produces more and more data, such as various user characteristics, how to use These features to build a unified Top-N recommendation model is a key issue; 3. In practice, it is difficult to combine and measure all features of items to produce HIN recommendations

Considering all the features may take a lot of time and lead to the problem of overfitting. Therefore, it is necessary to choose features reasonably in heterogeneous information networks.

Images