Abstract Given the frequent cost overruns and schedule delays associated with tunnel construction projects, it is imperative that a detailed estimation of both be developed and considered prior to starting construction. To this end, two artificial intelligence tools of Gaussian Process Regression (GPR) and Support Vector Regression (SVR) were used to forecast geology, construction time and construction costs of a road tunnel project. The initial training datasets applied in the prediction tools were accessed from the previously-constructed road tunnels and the pre-existing observations of the tunnel under consideration. Also, during the tunnel construction, more training datasets obtained in the constructed parts were added to the previous datasets and the pre-constructed predictions of the GPR and SVR tools were updated. Lastly, comparing the predictions made by the GPR and SVR tools with the actual mode of the tunnel, and comparing the pre-updating predictions with the post-updating ones, it was concluded that, the GPR and SVR tools have presented very good predictions and they have reduced the uncertainties regarding geology and construction time and costs to an acceptable level. But, the GPR tool has presented more accurate results than the SVR tool. Also, the updating procedure can significantly increase the predictions accuracy.

中文翻译：

---

使用人工智能工具进行隧道掘进决策

摘要 考虑到与隧道建设项目相关的频繁的成本超支和进度延误，在开始建设之前必须对两者进行详细的估算。为此，利用高斯过程回归（GPR）和支持向量回归（SVR）两种人工智能工具对某公路隧道项目的地质、施工时间和施工成本进行预测。应用在预测工具中的初始训练数据集是从先前建造的公路隧道和正在考虑的隧道的预先存在的观察中访问的。此外，在隧道施工过程中，更多在施工部分获得的训练数据集被添加到以前的数据集中，并更新了探地雷达和 SVR 工具的预先构建的预测。最后，将探地雷达和 SVR 工具的预测与隧道的实际模式进行比较，并将更新前的预测与更新后的预测进行比较，得出的结论是，探地雷达和 SVR 工具提供了非常好的预测，并且将地质、施工时间和成本的不确定性降低到可接受的水平。但是，GPR 工具比 SVR 工具提供了更准确的结果。此外，更新程序可以显着提高预测精度。但是，GPR 工具比 SVR 工具提供了更准确的结果。此外，更新程序可以显着提高预测精度。但是，GPR 工具比 SVR 工具提供了更准确的结果。此外，更新程序可以显着提高预测精度。