基于ARIMA和LSTM模型的建筑安全事故预测

Prediction of Construction Safety Accidents based on ARIMA and LSTM Models

研究比较差分自回归移动平均模型(Autoregressive Integrated Moving Average model,简称ARIMA)与长短期记忆神经网络(Long Short Term Memory,LSTM)模型在建筑安全事故预测中的效果。采用2012—2018年全国建筑安全事故快报数据训练ARIMA及LSTM模型,并对全国每年、每月发生的建筑安全事故次数进行预测,使用RMSE和MAE作为评价指标对比两种模型的预测准确率。ARIMA(1,1,0)模型和LSTM模型的RMSE、MAE值分别为8.1318、6.5911和16.4341、14.5534。结果表明,ARIMA模型比LSTM模型更适于预测建筑安全事故发生次数。

This paper proposes to study and compare the construction safety accident prediction effect of the Autoregressive Integrated Moving Average(ARIMA) model and the Long Short Term Memory(LSTM) model. The ARIMA and LSTM models are trained with the datasets of the national construction safety accident report from 2012 to 2018, and the number of construction safety accidents that occur every year and every month in China is predicted. RMSE and MAE are used as evaluation indicators to compare the prediction accuracy of the two models. The RMSE and MAE values of ARIMA(1,1,0) model and LSTM model are 8.1318, 6.5911 and 16.4341, 14.5534, respectively. The results show that ARIMA model is more suitable for predicting the number of construction safety accidents than LSTM model.