FastGCN recommendation-based sample feature aggregation method

Abstract

The invention discloses a FastGCN recommendation-based sample feature aggregation method. The method comprises the following steps of: 1, firstly, determining all node objects in a current network; 2, obtaining a high-order degree matrix of the feature matrix; 3, aggregating the obtained high-order degree matrixes into a degree matrix in a final form; and 4, performing FastGCN network recommendation by using the processed feature sample, and obtaining recommendation content with higher quality, thereby improving efficient operation of data service business of a company. The method has the beneficial effects that the fusion of the FastGCN model sample attribute features is completed by using a technical means of local feature fusion. The fused sample objects have richer attribute features, and the categories of the objects can be distinguished more easily when the neural network performs feature mapping on the objects, so that the recommendation precision of the model is improved. Application experience is improved.

Description

technical field [0001] The present invention relates to a sample feature aggregation method, in particular to a sample feature aggregation method based on FastGCN recommendation. Background technique [0002] At present, the application scope of the recommendation system model based on FastGCN is gradually expanding, and the model speeds up its recommendation speed on the basis of GCN. However, when it quantifies the characteristics of the original data, it only collects the explicit information part of the map (hereinafter referred to as graph or network), while ignoring the implicit information in these data. For example, in an encyclopedia index network, an entry is a dictionary, which will list reference books related or similar to certain features of the dictionary as its extended reading. But in the sample data of the FastGCN model, objects like dictionaries lack the features of other objects related to them. In fact, however, they are likely to be similar in some fe...

AI Technical Summary

Problems solved by technology

But in the sample data of the FastGCN model, objects like dictionaries lack the features of other objects related to them

However, in fact they are likely to be similar in some characteristics, but the current object's performance in this characteristic is weaker

This shows that the model has a technical flaw in data processing, which makes the model degrade the user's application experience in actual use

And it is not conducive to the healthy operation of the company's big data business

Images