The invention discloses a news click bait detection method based on network migration, which comprises the following specific implementation steps of: 1, when one sentence or one news arrives, preprocessing the sentence or the news; 2, word vectors are pre-trained, representation learning is conducted through a large number of corpora, and word vector representation of each word is obtained; 3, training is carried out in the source field, and an optimal effect of the model in the source field is achieved by training the convolutional neural network and the parameter model of the adaptation layer; 4, training is carried out in the target field, and the model can utilize the priori knowledge of the source field through the micro-adjustment adaptation layer; and 5, classification: when a piece of news comes, performing word segmentation preprocessing. The transfer learning model can achieve similar performance on the target language with less annotations, the effectiveness and robustnessof the model are indicated, and the problem that time and labor are wasted when a bait data set is clicked through manual annotation is solved.