Social robot account detection method based on feature evaluation and improved algorithm

Abstract

The invention discloses a social robot account detection method based on feature evaluation and an improved algorithm, and the method comprises the steps: selecting nine feature attributes as data forming a data set from the point of view of the dynamic change of a social account, and carrying out the classification evaluation of indexes through employing an AUC decision tree on the basis of introducing a machine learning random forest algorithm to recognize the social robot account, and selecting a plurality of decision trees with better performance, and voting to generate model output according to generated results. According to the method, the multi-dimensional dynamic feature social robot account detection model is constructed, the random forest optimization algorithm based on the AUC evaluation standard is combined, low-comprehensive-cost and high-accuracy detection of the social robot accounts is achieved, and the accuracy, the precision ratio and the recall ratio generated by the matrix are confused through the classification algorithm to serve as three indexes for verifying the effectiveness of the detection method.

Description

technical field [0001] The invention belongs to the technical field of network security and automation, and relates to a social robot account detection method based on feature evaluation and an improved algorithm. Background technique [0002] A social robot refers to a program that runs autonomously, publishes generated content, and imitates human activities in order to achieve a specific purpose in a social network. In recent years, as social media platforms have become the most influential public opinion field in the current society, social robots have gradually realized the transformation from the role of information dissemination intermediary to the main role of information dissemination. Studies have shown that a large number of social media users are currently not real humans, but social robots manipulated by automated programs. In addition, the proportion of information generated and published by these robot accounts in the total platform traffic is increasing signif...

AI Technical Summary

Problems solved by technology

However, the above-mentioned social robot detection algorithm mostly focuses on account attribute features in feature selection, and the model structure is single and the accuracy rate is low, which has certain shortcomings.

Images