With over 80 billion high-resolution photos collected by Street View cars, analyzing these images manually would have been impossible; instead, Google’s finely-tuned machine learning algorithms automatically extract information from geo-located images. The algorithms help the app extract street names and house numbers from photos taken by Google’s “Street View” cars and increase the accuracy of search results. For example, according to the Google Research Blog, the company introduced machine learning to Google Maps, improving the usability of the service. In some machine learning applications, computers are initially programmed to learn how to solve problems, but can change and improve algorithms on their own—making faster, more accurate predictions. We’ve all heard the buzz words by now, but just what is machine learning in construction?
This improvement demonstrates the critical role of NOI in the model’s predictive power and supports the study’s findings on the strong connection between SOI and NOI. Similarly, proper PPE usage and Relevant Safety Training positively contribute to mitigating incident severity, emphasizing the actionable insights these features provide for safety interventions. Figure 10 presents the SHAP summary plot, ranking features by their average absolute SHAP values, with higher values indicating greater influence. SHAP values provide a robust method to determine the impact of each feature on the model’s output, ensuring transparency and actionable insights for stakeholders31,52. To complement these rankings, SHAP analysis was performed to enhance the interpretability of the ML models and quantify the influence of individual features on the prediction of safety incident severity. The strong role of precipitation indicates that adverse weather conditions substantially affect severity outcomes, as heavy rainfall can compromise visibility, increase slip risks, and weaken scaffolding stability.
In contrast, KNN and SVM were the least effective, struggling to differentiate between severity levels due to their inability to capture non-linear relationships and complex decision boundaries. It essentially helps to find the needle in a haystack https://cafelam.com/unleashing-the-magic-of-roller-backer-a-comprehensive-guide/ of data, taking in large quantities of complex data and identifying patterns to provide reliable, effective and repeatable results. Given the results, XGB is recommended as the most suitable model for implementing proactive safety assessment systems, given its superior performance in distinguishing incident severity and handling complex datasets. The area under the ROC curve (AUC-ROC) is a critical metric that quantifies the model’s ability to correctly classify incidents at different severity levels, with values closer to 1.0 indicating superior performance.
Predictable schedules
Road areas followed at 10%, where traffic accidents were the primary cause of fatalities. These statistics establish the quantitative profile of the dataset and complement the categorical distributions illustrated in Fig. The study categorizes safety incidents according to the Occupational Injury and Illness Classification System (OIICS)22. The dataset consists of construction-based safety incidents from Saudi Arabia, primarily from Makkah and Riyadh regions, covering incidents from 2018 to 2024. The second phase of the study involved developing comprehensive ML models to predict the nature (NOI) and severity (SOI) of construction incidents, providing a useful framework for preventing future accidents.
Baker et al.19 leveraged natural-language features and model ensembles (e.g., RF, XGB, linear SVM, stacking) for injury classification over a large repository of narrative reports. They reported encouraging accuracy but offered limited analysis of which factors most strongly drive fatal outcomes. While influential, this study omitted injury-level variables, a limitation that may constrain predictive fidelity16. Recent work has https://open-innovation-projects.org/blog/use-free-and-open-source-software-to-effortlessly-open-dwg-files begun to apply ML to construction safety, moving beyond conventional post-hoc or purely descriptive analyses toward predictive modelling of incident outcomes14,15.
AI can scan the signals you already generate across construction projects, such as production rates, submittals, change activity, procurement lead times, and rework trends. It learns patterns from your past jobs and compares them to what’s happening now, across a broad range of signals like RFIs/submittals cycling, long-lead items drifting, inspection rework, or productivity falling below baseline. AI is already showing up across a broad range of construction projects. Everything from drones and robotics to machine learning and virtual tools helps teams compare plans and coordinate changes faster, which comes with many benefits. You may be thinking AI sounds out of place in a hands-on industry, but McKinsey reports that large construction projects typically run 20% longer than scheduled and up to 80% over budget.
The practical applications of machine learning in construction
- The dataset contains non-linear relationships and interactions between explanatory variables (e.g., the combined effect of PPE usage, safety training, and incident location), which are better captured by ensemble methods.
- While this study focused on the Saudi Arabian construction industry, its methodology can be adapted for other regions by incorporating local climate, workforce demographics, and regulatory frameworks.
- You capture 360° footage, it recognizes what’s installed, then it’s brutally honest about what’s missing, where, and how that maps back to the plan.
- Boosting algorithms, for example, iteratively refine the model by focusing on hard-to-classify instances, ensuring improved performance on imbalanced and noisy datasets.
- Table 1 provides a concise synthesis of recent ML-based studies on construction-incident prediction.
The lowest Micro-Average AUC scores were observed for KNN (0.68) and SVM (0.62), suggesting that these models struggled to effectively differentiate between different severity levels. While RF benefited from its ensemble learning capability, it still struggled in cases where severity levels overlapped. GB followed closely with a Micro-Average AUC of 0.86, reinforcing the effectiveness of boosting-based techniques in handling complex classification tasks.