Time Series Prediction Models for Assisting the Diagnosis and Treatment of Gouty Arthritis

Clinical gout arthritis data tracks changes as essential indicators and reflects the recurrence status of patients within several weeks after patients’ medication. Although the data may contain rich patient information, it is difficult to be fully utilized due to clinical data quality issues such as various time lengths, data missing, irregular sampling, etc. Time series prediction models have the potential to deal with these data problems. This paper compares a list of time series prediction models on the indicators of patients with gouty arthritis. We collected real data from the Guangdong Provincial Traditional Chinese Medicine Hospital including 160 patients. The Bidirectional long short-term memory (Bi-LSTM) model and the Crossformer model are applied to predict future physiological indicators and the recurrence status of patients. According to the results of Bi-LSTM and Crossformer, time series prediction models demonstrate strong performance in forecasting physiological indicators and the recurrence status of patients.