Characteristics and preparation of a polarization beam splitter based on a chalcogenide dual-core photonic crystal fiber.

We reported on a polarization beam splitter based on a novel chalcogenide dual-core photonic crystal fiber. The glass matrix of the optical fiber is Ge10As22Se68. We used computerized numerical control precision drilling methods to manufacture preforms. Then the preform was drawn into an optical fiber with a regular hole structure. The maximum extinction ratio reached -32.76 dB with a 26.27 mm-long optical fiber. Numerical results show that the shortest working length of the designed polarization beam splitter is 636 µm. In addition, the modeling analysis based on the actual structure shows that the theoretical value is consistent with the measured value.

Predictive Analysis of Hospital HIS System Usage Satisfaction Based on Machine Learning.

Hospital information system (HIS) can provide a full range of information support for various hospital business activities and information collection, processing, and transmission, helping medical service providers. And HIS can reduce medical service costs and improve work efficiency, greatly reducing errors in diagnosis and treatment. Although the advantages of using the HIS are obvious, there are still some challenges in its use, the most prominent being how to make the medical staff use HIS effectively. Based on this background, this paper uses machine learning (ML) technology to predict and analyze the satisfaction of HIS use in hospitals and completes the following work: firstly, introduce the situation and development trend of HIS construction at home and abroad and provide theoretical basis for model design. The related development technologies are discussed and studied in detail. Second, the ML algorithm is used to provide a prediction strategy. The support vector machine (SVM) can handle small data sets well, and this study applies the AdaBoost technique to improve the model's generalization ability and accuracy. Lastly, a diversity metric is included to guarantee that the basic learner has good variety in order to increase the algorithm's performance. Accuracy rates may reach more than 95% in the case of tiny data sets, according to the self-built data set used for testing. This proves the superiority of the model proposed in this paper.