Streaming Machine Learning For Real-Time Gas Concentration Prediction
Haibo Wu, Shi Shi-liang, Qifeng Nian · 2019
The monitoring data of the gas in coal mines are characterized by the streaming big data, because of the development of the Internet of Things. To accurately and dynamically predict the gas concentration, and improve the accuracy of the risk prediction of gas outburst, first, we established a model based on the streaming data to predict the gas concentration, using the streaming machine learning algorithm. This model was proposed based on the principal component analysis and the streaming linear regression method. In addition, we proposed a prototype system which supports iterative prediction model refreshment for live data streams for the real-time prediction of gas concentration, using the Spark Streaming. Moreover, the experimental results show that the update cycle of the model was 45s. Therefore, this system, based on the streaming machine learning can properly predict the real-time gas concentration.