Hyperthyroidism Progress Prediction with Enhanced LSTM.

In this work, we propose a method to predict the progress of the hyperthyroidism disease based on the sequence of the patient’s blood test data in the early stage. Long-Short-Term-Memory (LSTM) network is employed to process the sequence information in the tests. We design an adaptive loss function for the LSTM learning. We set bigger weights to the blood test data samples which are nearby the range boundaries when judging the hyperthyroidism. We have carried out a set of experiments against a real world dataset from a hospital in Shanghai, China. The experimental results show that our method outperforms the traditional LSTM significantly.