A Semantic Path Based Approach to Match Subgraphs from Large Financial Knowledge Graph

摘要

In the past, people studied the stock market based on the assumption that the stock entity is known to be affected by the news. However, due to this assumption, these methods inevitably ignore the news without stock entities, and many news without stock entities will also have a significant impact on financial markets. In order to solve this problem, this paper proposes a subgraph matching algorithm based on semantic paths. Matching subgraphs on a knowledge graph that collects a large amount of stock market information and matching the affected stock entities from the semantic level can make a comprehensive analysis on Various news with or without entities. The main research work and achievements of this paper are as follows: First, starting with structured data, the paper complements semi-structured data and unstructured data to build a knowledge graph of the stock market and covering most of the stock market entities. Secondly, based on the analysis of LDA topic model, this dissertation extracts useful topics from financial news and constructs a news graph. A subgraph matching algorithm based on semantic path is proposed. From the knowledge graph, subgraphs matching with news graph are searched for mining the associated entities in the financial news. Finally, according to the result of subgraph matching, experiments and simulated investment are designed. The strategy achieved 15.96% excess return relative to the benchmark. The effectiveness of the subgraph matching algorithm based on semantic path is verified, and the feasibility of the algorithm in actual investment is proved.