Comparison and Evaluation of Multi-Index Prediction from the Perspective of Network Topology and Latent Semantics^*

In the research field of technological innovation mining and prediction, finding the opportunity of technological innovation and understanding the trend of technological development are hotspots. This paper provides the method of technological trend mining, which is conducive to improving the efficiency and benefit of the existing mining methods and promoting innovation institutions to enhance their technological competitiveness. Taking the topological structure of industrial patent data and the set of patent text as the research object, the dominant and recessive characteristics of the technology field are mined through the multi-method system such as LDA, Apriori association rule and link prediction, and the technology development trend of the target industry is analyzed. The empirical analysis of 10 years of patent data in the field of driverless vehicles shows that the composite index combining potential semantic technical topic association and dominant topology structure has a relatively good prediction effect under multiple indicators, and has certain validity and reliability for the technology fusion trend and analysis of the target industry.