Incremental tensor decomposition method and system for open source event association prediction

Abstract

The invention discloses an incremental tensor decomposition method and system for open source event association prediction, and the method comprises the steps: building an event global index structurethrough employing a multi-dimensional feature imported by event data as a core element category, carrying out the incremental maintenance of the event global index structure according to new arrivalevent data, and completing the construction of an event index incremental data structure; adopting an implicit factor matrix representation structure to construct and initialize an event tensor decomposition model, establishing a tensor incremental updating rule according to a random gradient descent method, and updating the tensor decomposition model through dynamic event data; predicting the incidence relation between any two event elements and the incidence relation between any two events according to the implicit factor matrix, and extracting event record principal components. The method and system are used for solving the problems of event data high-order sparseness, event principal component exploration and event data continuous updating in the prior art, realizing adaptation to high-order sparseness, supporting principal component exploration and supporting data continuous updating.

Description

technical field [0001] The invention relates to the technical field of data processing, in particular to an incremental tensor decomposition method and system for global open source event correlation prediction and principal component extraction. Background technique [0002] With the continuous development of big data technology, international relationship event data is constantly enriched and improved. Among them, the open source event data represented by GDELT is particularly well-known for providing massive event information involving different geographical regions and different time periods around the world. In recent years, the monthly data subset of GDELT can reach the scale of tens of millions of events; the annual data scale of GDELT is in the order of tens of millions or even hundreds of millions. Mining and utilizing the internal structure and correlation between GDELT event data has become an important way to accurately study and judge complex international and d...

AI Technical Summary

Problems solved by technology

[0007] The present invention provides an incremental tensor decomposition method and system for open source event association prediction, which is used to overcome the defects in the prior art that many activities between entities cannot be directly observed due to the high-order scarcity of event data, and realize adaptation to high-order sparse, Support principal component exploration and continuous update of data

Images