OO-LSTM: A trusted medical transfers prediction model with on-chain and off-chain data fusion

When the medical services surrounding patients cannot meet the needs of patients, transfer treatment has become an unavoidable part in the current medical environment. Initially, people choose the transfer path within their own cognitive range. With the development of the Internet, people can obtain the transfer paths of patients similar to them from all over the country through the Internet for reference, further combine machine learning models such as RNN, LSTM, and CNN to recommend the best transfer path. However, the treatment behaviors usually span multiple medical institutions, and it is difficult to comprehensively and efficiently share the cases, examination results and treatment process between the institutions. In addition, the source of traditional prediction model’s data set is opaque, and the integrity of data needs to be verified, which all affect the accuracy of prediction result. In order to solve the above problems, we improve the traditional LSTM, and propose OO-LSTM which integrates data sets on the blockchain and off the blockchain, uses the analysis method of transfer path based on blockchain association and traceability mechanism, and supplements and cross-validates the data on the chain and off the chain. Based on OO-LSTM, we can obtain a reliable and complete prediction data set and provide more accurate and credible recommendations for patients according to the conditions of patients in the current medical institutions. Experiments have proved that our method has higher credibility and more accurate results than traditional prediction models at the same cost.