Accuracy metrics have been widely used for the evaluation of predictions in machine learning. However, the selection of an appropriate accuracy metric for the evaluation of a specific prediction has not yet been specified. In this study, seven of the most used accuracy metrics in machine learning were summarized, and both their advantages and disadvantages were studied. To achieve this, the acoustic emission data of damage locations were collected from a pile hit test. A backpropagation artificial neural network prediction model for damage locations was trained with acoustic emission data using six different training algorithms, and the prediction accuracies of six algorithms were evaluated using seven different accuracy metrics. Test results showed that the training algorithm of “TRAINGLM” exhibited the best performance for predicting damage locations in deep piles. Subsequently, the artificial neural networks were trained using three different datasets collected from three acoustic emission sensor groups, and the prediction accuracies of three models were evaluated with the seven different accuracy metrics. The test results showed that the dataset collected from the pile body-installed sensors group exhibited the highest accuracy for predicting damage locations in deep piles. Subsequently, the correlations between the seven accuracy metrics and the sensitivity of each accuracy metrics were discussed based on the analysis results. Eventually, a novel selection method for an appropriate accuracy metric to evaluate the accuracy of specific predictions was proposed. This novel method is useful to select an appropriate accuracy metric for wide predictions, especially in the engineering field.

The monitoring and prediction of slope displacement plays a vital role in slope stability analysis. The displacement changes influences the stability of the slope. This paper described an approach to monitoring and predicting the variation of slope displacement. Firstly, The target slope was photographed from multiple angles by unmanned aerial vehicle (UAV) gimbal controller. Then an image algorithm of visual motion is applied to reconstruct the point cloud of the slope, and to get the digital elevation model (DEM) of the slope. The DEM model was used to calculate the displacement of the monitoring area. The Displacement information measured by UAV photogrammetry will be divided into training sample set and the prediction sample set. Using the AFSA-Elman algorithm neural network, the displacement sequences were trained and predict the variation of the displacement sequence. Compared with the on-site measurement results and existing Elman network, the result demonstrated that the UAV measurement technique have a high efficiency in monitoring the displacement at each measured point of the slope. And the AFSA-Elman network was proved a higher precision and better convergence comparing with the traditional Elman network, which was suitable for the predict the displacement of the key measuring point in the slope.

The construction industry is undergoing a transformative shift through automation, with advancements in Generative AI (GenAI) and prompt engineering enhancing safety and efficiency, particularly in high-risk fields like underground construction, geotechnics, and mining. In underground construction, GenAI-powered prompts are revolutionizing practices by enabling a shift from reactive to predictive approaches, leading to advancements in design, project planning, and site management. This study explores the use of Google Gemini, a recent advancement in GenAI, for the prediction of rockburst intensity levels in underground construction. The Python programming language and the Google Gemini tool are combined with prompt engineering to generate prompts that incorporate essential variables related to rockburst. A comprehensive database of 93 documented rockburst cases is compiled. Subsequently, a systematic method is established that involves the categorization of intensity levels through data visualization and factor analysis in order to identify a reduced number of unobservable underlying factors. Furthermore, K-means clustering is utilized to identify data patterns. The gradient boosting classifier is then employed to predict the intensity levels of rockburst. The results demonstrate that GenAI and prompt engineering offers an effective approach for accurately predicting rockburst events, achieving an accuracy rate of 89 percent. Through predictive modeling with GenAI, construction engineering experts can proactively evaluate the likelihood of rockburst, allowing for improved risk management, optimized excavation strategies, and enhanced safety protocols. This approach enables the automation of complex analyses and provides a powerful tool for real-time decision-making and predictive insights, offering significant benefits to industries reliant on underground construction. However, despite the considerable potential of GenAI and prompt engineering in the construction sector, challenges related to output accuracy, the dynamic nature of projects, and the need for human oversight must be carefully addressed to ensure effective implementation.