Risk research and judgment method based on multi-modal learning and LSTM

Abstract

The invention discloses a risk research and judgment method based on multi-modal learning and LSTM, and the method comprises the following steps: obtaining related data which requires space-time information features and other multi-source data information; carrying out preprocessing and multi-modal information fusion on the data; performing data processing according to the proposed data stream togenerate intermediate result data; performing risk research and judgment based on the improved LSTM; and obtaining a risk research and judgment result and calculating the accuracy according to the verification set data. According to the invention, through a multi-modal information fusion technology, multi-source data is subjected to technical fusion, and a vector with a specific length is generated. At present, an LSTM model is excellently transformed and innovated to a certain extent, an existing method for carrying out risk research and judgment on special personnel through a manual or simple data statistics method is optimized, and the risk research and judgment method for the special personnel in the public safety field under the data background is provided.

Description

technical field [0001] With the development of the economy and digital society, the risk assessment of special personnel has always been an important research topic in the field of public security, and plays an important role in maintaining social harmony and stability. Portraits and risk analysis of special personnel in smart cities play an important role in the field of public security. Under such a background, the present invention proposes a risk research and judgment method based on multi-modal learning and LSTM. Background technique [0002] Many data are graph structures, such as social networks, economic networks, biological networks, information networks (Internet websites, academic citations), the Internet, and neural networks. The network is their common language, so it has great research value. A potential machine learning task can be done. [0003] Recent advances in representation learning embedding vectors have led to a proliferation of methods for learning...

AI Technical Summary

Problems solved by technology

[0003] Recent advances in representation learning embedding vectors have led to a proliferation of methods for learning continuous-space representations of graphs

Images