Navy detection model construction method and system and navy detection method

Abstract

The invention relates to a navy detection model construction method. The navy detection model construction method comprises the steps of (1) conducting task segmentation on a set of sample data to obtain a plurality tasks, and extracting average features to obtain a training sample set of the tasks; (2) selecting the features of the tasks to obtain a feature weight matrix of the tasks; (3) setting a threshold value delta, judging whether the maximum value of one column vector in the feature weight matrix is larger than the threshold value delta or not, and if yes, executing the step (4); if not, abandoning the column vector, and executing the step (5); (4) adding the column vector into a sharing feature item set; (5) judging whether columns vectors which are not compared with the threshold value delta exit in the feature weight matrix or not, and if yes, executing the step (3); if not, executing the step (6); (6) inputting a new training data set; (7) obtaining a linear classification value through calculation; (8) setting a navy threshold value, and determining that data are from a navy when the linear classification value is larger than the navy threshold value. According to the navy detection model construction method, a navy detection model is built through a multi-task learning method, so that a navy user is conveniently and rapidly recognized.

Description

technical field [0001] The invention relates to a water army detection method, in particular to a water army detection model construction method and system based on multi-task learning and a water army detection method, belonging to the field of network security. Background technique [0002] With the popularity of social networks, forums have become one of the most popular online applications. However, the open nature of online forums determines that it is difficult to strictly supervise the information in the forums, which has led to the emergence of a group of online trolls who deliberately spread certain remarks for the purpose of profit. From the "July 23" bullet train accident sky-high compensation incident to the Qin Huohuo incident, cyber trolls have had a serious impact on the network environment and even social order. It can be seen that the identification and supervision of cyber trolls is imminent. [0003] The traditional machine learning method for navy detect...

AI Technical Summary

Problems solved by technology

There are several disadvantages in this learning method: First, the samples in the training set only reflect the information of a single task, and the learning results are often limited by specific tasks

For example, Sina Weibo’s standards for judging trolls are different from those of forums. The troll judgment model learned from Sina Weibo’s rules may be misjudged as trolls when applied to other forums. user

Second, when the number of samples in the training set is small, the model constructed by single-task learning has a certain loss of information characteristics.

The traditional method tries to solve this problem by generating new samples through oversampling technology, but the generated new samples still cannot guarantee that the characteristic attributes of the original samples are completely consistent.

Images