Litcius/Paper detail

Multivariate Time Series Forecast in Industrial Process Based on XGBoost and GRU

Naiju Zhai, Peifu Yao, Xiaofeng Zhou

202042 citationsDOI

Abstract

In this paper, a time series prediction model that merges eXtreme Gradient Boosting (XGBoost) and Gate Recurrent Unit (GRU), XGB-GRU model, is proposed for multivariate time series prediction in industry. The XGB-GRU model uses XGBoost's strong feature extraction capabilities to extract the hidden information of multiple control variables in industrial data. Next, the model uses GRU's unique gating unit to extract the timing information in the industrial data. Finally, the importance of XGBoost output variables to guide actual production and solve the problem of inexplicability of neural networks. Predicting the temperature of the heating furnace verifies that the proposed XGB-GRU is better than a single XGBoost and GRU model. And the model has a good fit to the predicted value.

Topics & Concepts

Multivariate statisticsTime seriesComputer scienceExtreme learning machineBoosting (machine learning)Artificial intelligenceData miningGradient boostingProcess (computing)Artificial neural networkFeature extractionMachine learningOperating systemRandom forestNeural Networks and ApplicationsFault Detection and Control SystemsImage Processing and 3D Reconstruction
Multivariate Time Series Forecast in Industrial Process Based on XGBoost and GRU | Litcius