The invention provides a news popularity prediction model training method based on a deep neural network, and the method comprises the steps: obtaining news article data in a specific theme setting time period, carrying out the data cleaning through Pands, carrying out the sequential grouping according to a set time length, and obtaining a news popularity sequence obtained through the time sequence arrangement; according to the news popularity sequence, sequentially taking a continuous sequence with a sampling length of w as an input sample from a first popularity, and sampling data of a nextperiod as an output sample to construct a training sample set; randomly selecting training samples from the training sample set to train the news popularity prediction model based on the LSTM network,making correlation analysis by adopting Pearson correlation coefficients to delete bad training samples, and cycling the training process until training concludes. According to the invention, the news popularity prediction model used for predicting the non-trend, non-seasonal and non-linear news popularity with high accuracy can be obtained.