基于机器学习的英语翻译机器人异常智能诊断模型构建及仿真  

Construction and Simulation of abnormal intelligent diagnosis model for English translation robot based on machine learning

作　　者：刘佳[1] LIU Jia(Xi’an FanYi University,Xi’an 71015,China)

机构地区：[1]西安翻译学院,西安710105

出　　处：《自动化与仪器仪表》2023年第1期241-245,共5页Automation & Instrumentation

基　　金：2022陕西省哲学社会科学重大理论与现实问题研究项目《传播学视阈下的终南文化术语英译研究》(2022ND0168);2022西安市社会科学规划基金项目《“中国梦”在陕西乡土文学中的历史探索与启示》(22LW192)。

摘　　要：针对英语翻译机器人在出现故障时诊断反馈速度慢且准确度不高的问题,基于机器学习的方法提出一种异常智能诊断模型。仿真实验结果表明,加入动态时间步改进方法以及RF-GBDT特征选择算法的LSTM神经网络故障诊断模型在进行异常诊断时反馈速度快且准确度得到了提高。相较于未经特征的故障诊断模型,基于RF-GBDT特征选择算法的故障诊断模型在进行特征选择时准确率均得到了提升,而经过动态时间步改进的LSTM神经网络故障诊断模型相比于未改进前,数据分类准确度提高到了93%,每帧序列数据计算时间步长减少了39%,经过最后的故障诊断系统测试,证明了设计的诊断模型具体有较高的准确性和实用性。In order to solve the problem that the feedback speed of English translation robot diagnosis is slow and the accuracy is not high,an abnormal intelligent diagnosis model based on machine learning is proposed.The simulation results show that the fault diagnosis model of LSTM neural network with dynamic time step improvement method and rf-gbdt feature selection algorithm has fast feedback speed and improved accuracy in abnormal diagnosis.Compared with the fault diagnosis model without features,the fault diagnosis model based on rf-gbdt feature selection algorithm has improved the accuracy of feature selection.Compared with the fault diagnosis model without features,the LSTM neural network fault diagnosis model improved by dynamic time steps has improved the accuracy of data classification to 93%and the calculation time step of each frame sequence data has been reduced by 39%.After the final test of the fault diagnosis system,It is proved that the designed diagnosis model has high accuracy and practicability.

关 键 词：机器学习 故障诊断 随机森林算法 梯度提升度算法 LSTM神经网络 

分 类 号：TP392[自动化与计算机技术—计算机应用技术]