Network representation learning method based on random walk

Abstract

The invention provides a network representation learning method based on random walk, and the method comprises the following steps: building an NSRW model, wherein the NSRW model is represented as a function G = (V, E), and V = {v1, v2,... Vn} represents a node and is used for representing an entity in a network, E = {e1, e2,... En} represents an edge and is used for representing a relationship between entities in the network; calculating the similarity of two adjacent nodes in the network, wherein the similarity calculation formula of the two adjacent nodes is as follows: (shown in the description), nab represents the number of common neighbors between the node va and the node vb, and ka and kb respectively represent the node degrees of the node va and the node vb; calculating a walk probability between adjacent nodes according to the similarity; performing random walk according to the walk probability to generate a node sequence; and according to the node sequence, performing representation learning of the nodes to obtain low-dimensional representation of the nodes. The network representation learning method provided by the invention is more accurate in classification.

Description

【Technical field】 [0001] The invention relates to the field of network representation learning, in particular to a random walk-based network representation learning method. 【Background technique】 [0002] Network node classification is a major task in the field of network analysis, and there have been many research results, such as the combination of support vector machine (Support Vector Machine, SVM) and rule-based classifier (Rule-based classifier, RBC), decision tree and CRFs joint optimization model and semi-supervised network classification method based on random graph, etc. However, most of these methods focus on improving classification results using approximate inference, which is difficult to deal with network sparsity. [0003] Network representation learning (NRL) provides an effective way to solve the above problems. NRL converts network nodes into low-dimensional real-valued vectors and preserves network topology to the greatest extent. After obtaining low-di...

AI Technical Summary

Problems solved by technology

[0004] In related technologies, the DeepWalk algorithm is a commonly used algorithm in the NRL algorithm. It uses random walks to generate node sequences and obtains the vector representation of each vertex in the network through the Skip-Gram model; however, this algorithm ignores the different node links. Similarity, in the sampling process, the walking probability of any node is the same, making the classification accuracy not high

Images