A symbol prediction method and system based on bridging domain transfer learning

Abstract

The invention discloses a symbol prediction method and system based on bridging domain transfer learning. According to the prediction method, the bridging domain is used, so that the network without intersection generates a public knowledge space, and knowledge of the symbolic network is efficiently migrated to the target network without symbols. According to the prediction method, the bridging domain selection algorithm based on the status theory is high in universality, and the target domain can be predicted without any bridging domain symbol information. According to the prediction method,interference samples can be effectively removed, so that effective migration of reliable knowledge is ensured, the prediction error is small, and symbol information in the target network can be accurately predicted.

Description

technical field [0001] The invention relates to the field of network information processing, in particular to a symbol prediction method and system based on bridging domain migration learning. Background technique [0002] The task of the symbol prediction problem is to predict the symbols linked in a symbol network. A signed network refers to a network in which the edges in the network have signs, where a positive sign indicates a positive relationship between users, and a negative sign indicates a negative relationship between users. Research on the positive and negative prediction of links in symbolic social networks, the results have very important application value for personalized recommendation of social networks, identification of abnormal nodes in the network, user clustering, etc. However, the cost of obtaining symbolic information marked by experts is high, so symbolic information cannot be fully obtained, or even not obtained at all, so the lack of data in symbo...

AI Technical Summary

Problems solved by technology

However, the selection of intermediate domains is still a practical problem to be solved. Due to the sparse and uneven distribution of symbols in social networks, TTL cannot perform symbol prediction in social networks.

The existing selection of instances by selecting useful instances can overcome the sparsity and imbalance of symbol distribution in social networks, but useful information will be lost in these discarded instances, and it is impossible to realize the diversity of social networks. predict

Images