Dangerous Chemicals Monitoring Method Based on Sensor Array and LightGBM-SR Model
The dangerous chemicals monitoring method using the sensor array and the LightGBM-SR model is studied.Multiple sensors are used to obtain the real-time laboratory safety monitoring data.Non-linear stochastic resonance(SR) model is used to adjust the raw monitoring data.ExtraTrees,XGBoost,KNN and LightGBM models are selected as pattern recognition models.The sensor array raw data and SR adjusted data are input into four pattern recognition models,respectively.The regression prediction is conducted based on the test set data.Results indicate that the raw sensor array monitoring data with the four models prestent low predicting accuracy.Non-linear SR adjusted data with the recoginition models have higher predicting accuracy.The accuracy of LightGBM-SR model is improved from 78.75％ to 98.34％.ExtraTrees-SR has the best stability but still needs for longer time.The generalization ability and stability of XG-Boost-SR and KNN-SR are accetable,but the average accuracy is relatively low.LightGBM-SR model presente higher average accuracy,which is more suitable for the application of dangerous chemicals monitoring.
dangerous chemicals monitoringtoxic gases leakagenon-linear modelLightGBM model
万方数据
维普
