【摘要】 针对传统医疗器械故障预测方法存在预测精度和效率不高的问题,提出一种基于深度学习的医疗器械故障预测方法。首先,结合文本卷积神经网络(Text Convolutional Neural Network,Text CNN)的局部特征提取性能和基于Transformer的双向编码器表示(Bidirectional Encoder Representations from Transformers,BERT)预训练模型的上下文依赖捕捉性能,构建基于BERT-Text CNN的医疗器械故障预测模型。实验结果表明,本文模型的精确率、召回率和F1值分别为96.74%、98.53%和0.972 8,预测时长仅为10.62 s,明显优于现有的双向长短期记忆网络(Bidirectional Long Short-Term Memory,Bi LSTM)预测模型、循环神经网络(Recurrent Neural Network,RNN)预测模型和生成对抗网络(Generative Adversarial Network,GAN)预测模型。将本文模型应用于搭建的医疗器械故障系统中后,系统预测精度和效率显著提高,满足医疗器械故障高精度、高效率的预测需求,具备可靠性和稳定性。更多还原

【Abstract】 To address the issues of low prediction accuracy and efficiency in traditional medical device fault prediction methods,this article propose a deep learning-based method for medical device fault prediction.Firstly,combining the local feature extraction performance of text convolutional neural network(TextCNN) with the context dependency capture performance of the pre-trained model based on bidirectional encoder representations from Transformers(BERT),a BERT-TextCNN-based medical device fault prediction model is constructed.Experimental results show that the proposed model achieves a precision of 96.74%,a recall of98.53%,and an F1 score of 0.9728,with a prediction time of only 10.62 s,significantly outperforming existing prediction models such as bidirectional long short-term memory(BiLSTM),recurrent neural network(RNN),and generative adversarial network(GAN).Applying the proposed model to a medical device fault prediction system significantly improves the system’s prediction accuracy and efficiency,meeting the high-precision and high-efficiency prediction requirements for medical device faults,and demonstrating reliability and stability.更多还原