Prediction research of cervical cancer clinical events based on recurrent neural network
Yufang Yan, Kui Zhao, Jilong Cao, Huimin Ma
Abstract
Through the research on the existing time series prediction technology, most researchers mainly make predictions based on clinical events, without thinking about whether the previous clinical process is standard or not. This paper proposes a two-stage prediction model, RNN-2-DT, based on word vector representation and integrated into the standardized judgment of the diagnosis and treatment process. The model mines standardized clinical mode which is a standard of clinical process, using the method of binary K-means. Meantime, the clinical events prediction model using gated recurrent units (GRU) based on recurrent neural network (RNN) is constructed. The experimental results indicate that, compared with the clinical processes no considering standardized judgments, our model’s recall rate and mean average precision are increased by 7.2% and 4.3%, respectively.