Microblog water army identification method based on multi-task learning

The invention discloses a microblog water army identification method based on multi-task learning. The method comprises the following steps: S1, carrying out data preprocessing on a user attention quantity, a fan quantity, gender, a microblog grade, whether authentication is carried out or not, an authentication category, sunshine credit, a member category, a member grade and a blog article content, and dividing tasks into t1, t2,... and t5 according to user authentication categories; sorting pi belongs to delta5, delta5 is a symmetric group of all arrangements on five elements, training a classifier for each task by using an SVM (Support Vector Machine) model, and calculating a weight corresponding to each task in the sorting pi; s2, traversing each pi in delta 5 by solving the average error research sequence pi, determining pi (i) by minimizing the upper limit of a corresponding item of a task which is not solved, and determining an optimal learning sequence; and S3, according to the optimal learning sequen