Node Failures of Heterogeneous Networks Link Based on Improved MSLP Algorithm

When a heterogeneous information network and a target meta-path are selected, it is necessary to use the information provided by the historical data of the heterogeneous network to predict whether there will be a target path instance or target link between the nodes at both ends of the target meta-path in the future. In this problem, many factors affect the formation of target path instances between the nodes to be tested. Through the analysis of historical data of heterogeneous networks, the main task of this paper is to predict whether there will be failure instances under the target meta-path between the nodes to be tested in the future network. First, Limit Length Meta-path Generating (LLMG) algorithm based on the backtracking strategy is proposed, which solves the shortcomings of the traditional methods that need to rely on prior knowledge to select the relevant meta-path. Then, the Heterogeneous Link Entropy with Time (HLE-T) algorithm that dynamically combines link entropy and time is proposed. Taking into account the different effects of different nodes in the path instance and the establishment time of the path instance, the effectiveness of the HLE-T algorithm is verified on the data set of the Modified Supervised Link Prediction (MSLP) academic network. Next, through further analysis of the characteristics of heterogeneous networks, the traditional link prediction model based on supervised learning is improved, and the MSLP algorithm is designed.