Research on Joint Forecasting Technology of Fire Gas Concentration Based on ARIMA-SVM

The concentration of fire gases plays a significant role in fire evacuation, proper handling, and environmental protection. This paper selects the ARIMA-SVM combined forecasting method. The first step is to use the ARIMA model to model and predict the linear trend of the time series data of fire gas concentrations; then, the SVM model is applied to perform nonlinear fitting on the residuals predicted by ARIMA to compensate for the errors of the ARIMA model. Finally, by summing the prediction results of both models, a combined forecast of short-term nonlinear disturbances and long-term linear trends is achieved. Experimental results show that the ARIMA-SVM model exhibits higher accuracy and stability in predicting key fire gases such as TVOC compared to the individual ARIMA model prediction. This method provides new insights for monitoring and early warning in fires and can be extended to the prediction of other complex environmental gases.