A Review of Data Mining Strategies by Data Type, with a Focus on Construction Processes and Health and Safety Management

1. Introduction

The activities attributable to the construction sector according to the International Labour Organisation (ILO) classification are as follows: i) building, including excavation and the construction, structural alteration, renovation, repair, maintenance (including cleaning and painting) and demolition of all types of buildings or structures; ii) civil engineering, including excavation and the construction, structural alteration, repair, maintenance and demolition of structures such as airports, docks, harbours, inland waterways, dams, river and avalanche and sea defence works, roads and highways, railways, bridges, tunnels, viaducts and works related to the provision of services such as communications, drainage, sewerage, water and energy supplies; and iii) the erection and dismantling of prefabricated buildings and structures, as well as the manufacturing of prefabricated elements on the construction site [30].

Construction safety research is numerous and motivated by the alarming rates of accident and fatalities, focusing on two perspectives: management and technology [77]. In general, workplace safety management is based on organisational and technological strategies. Construction safety standards and accident reduction are achieved through information and worker training, aiming to enhance the level of risks perception associated with the production process. However, the impact of traditional accident prevention strategies has been limited due to their reactive and regulatory nature [45,77]. A relevant aspect is the increased risks associated with the organisation and production goals of construction companies.

According to Razi et al. [51], Artificial Intelligence (AI), is a broad field of computer science concerned with the developing intelligent robots capable of performing tasks that traditionally require human intellect. AI plays a crucial role in assisting construction supervisors in minimizing accidents, supporting project efficiency, and significantly improving operational safety. Alongside the advancement of information and communication technology, various innovative technologies have been investigated to aid and improve on existing management-driven safety management practices. Besides the aid of technologies, new injury prevention strategies have been developed for the construction industry. The risk analysis method is one of them which are used in safety programs to improve safety performance. A relevant factor is the relationship between the type of construction project and the type of accident.

Data mining methods are applicable in various fields, dealing with different types of data and objectives. Studies focusing on DM techniques applied to construction safety date back to no later than 2014 [52]. The study has been developed as part of the 2022-2024 Inail's research program and the objective “Study of the effectiveness and efficiency indices related to innovative technologies aimed at preventing the risk of injury in highly variable work environments”. The protocol of review is led by the Preferred Reporting Items for Systematic Reviews and Meta-Analysis Protocols (PRISMA) [50]. Section 1 of the article introduces the background and objectives of the investigation. Section 2 describes the materials and methods, while Section 3 presents the results obtained from the applying cluster analysis, PCA, and meta-analysis. Section 4 offers an extensive discussion of the results, considering the current state of the art and our future goals. Finally, Section 5 summarises the salient results achieved in this review.

2. Materials and Methods

The set of articles published from 2014 to September 2023, which were useful for the purposes of this review, was extracted from Scopus [9,20] and ProQuest [15]. Authoritative sites on conferences in the field of computer science and DM and Management in Construction field were queried; however, only Web of Conference provided an eligible contribution for the purposes for our review.

2.1. Selection and Inclusion Criteria

All searches were conducted using a combination of subject headings and free-text terms. We focused exclusively on peer-reviewed articles, conference papers and book chapters. The topics included were: “Machine Learning,” “Work,” “Construction,” and “Risk,” accross the following subject areas: (i) Engineering, (ii) Social and Environmental Sciences, (iii) Computational Sciences. The criteria applied in the search strategy are defined in Table 1. The final search strategy was developed through several preliminary searches including (i) articles, (ii) conference papers, (iii) book chapters.

Figure 1 summarise the result of the PRISMA document selection process. The collected paper dataset includes the information on Authors, Title, Year of publication, Source of title, Volume, Issue, Number of Pages, Citation number, DOI, Affiliations Author information, Abstract, Keywords, Type of Publication, and further information. Three authors (AP, AB, and DB) independently reviewed the titles and abstracts to assess the eligibility of all studies. We applied PRISMA procedures and checklists [50] in identifying topic, datasets and keywords and filtering content according to abstract, assessing the eligibility of publications in the research scope. Further insights were made into the selected articles by conductingfull-text review and analysing the content for search purposes (see Figure 1) [23]. Disagreements were resolved by a fourth evaluator (ML) until consensus was reached between the authors. Only studies that met the eligibility criteria were included.

2.2. Risk of Bias for Selected Studies

The risk in non-randomised studies was assessed on the basis of the following biases: (1) due to confounding, (2) in the selection of the types of data in the study, (3) in the classification of the study objective, (4) due to missing data, (5) in the measurement of outcomes, (6) in the evaluation metrics and (7) in the selection of the reported outcome. Each individual study included was assessed as having a low, moderate, severe, and critical risk of bias. If critical information was missing for the assessment of risk of bias, these studies were considered devoid of information.

2.3. Data Quality and Items

The titles and abstracts of the identified studies were independently checked at two different points in time. Eligibility and inclusion criteria were initially assessed on a subset of 30 studies before searching all databases. Decisions were made by examining both the abstracts and the full texts. Only studies that were complete and met all inclusion criteria were included in the qualitative and quantitative synthesis. The information and data included in the papers obtained through the PRISMA method, were then included in the review.

2.3. Study Design

The scientific articles falling under the eligibility criteria of PRISMA were pre-processed to extract information suitable for the purpose of review. The 61 papers included in the review were categorized by 33 source titles and by publication year (Table 2). The bibliometric analysis involved a review of the global literature and geographic mapping worldwide (Figure 2). The cluster analysis (HC) was used to find the best aggregations between groups. By the Silhouette index, it was found the best degree of aggregation to be represented in a cluster plot based on correlations and variances (Figure 3 and Figure 4). Principal component analysis (PCA) was useful to find the correlation classes between the various parameters of the dataset in a simplified reading of the results. Through PCA, we reduced the items and obtained the extent of correlation between variables, methods, and components. The meta-analysis of this classes was useful to estimate the reliability of HC and PCA results and the odds ratios OR and confidence intervals CI of groups of items. Spatial data collection, analysis, classification, and bibliometric analysis were performed with VOS viewer [65], R and GIS software.

3. Results

3.1. Study Selection and Bibliometric Analysis

The search strategy, based on the PRISMA eligibility criteria, yielded 61 papers that were included in the review and categorized by 33 source titles, (Table 2) and publication year. Regarding the latter, there was an increasing trend in publication from 2014 to September 2023, where the articles recorded the following trend: 1 paper each in 2014 and 2015, 3 papers in 2016 and 2017, 4 papers in 2019, 6 in 2020, 3 papers in 2021, 17 papers in 2022 and 23 papers in 2023.

Figure 2 shows a map of the 61 scientific articles included in the review from 2014 to September 2023, by country of origin, worldwide.

3.2. Classes of Data and DM Methods

In a separate dataset, study objective, type of data under investigation, applied DM methods, applied DM type (supervised, unsupervised, other), validation metrics (if available), the DM method found to be most effective, number of rows and columns in the dataset used by the authors (if available). The 61 selected articles provided 202 observations, 91 DM methods (50 of which were considered the best method), 4 survey purposes, 3 field, 7 data types, 3 resource type (as detailed in Table 3 and Appendix A and Appendix B). Among study objectives, we obtained X1 classifying (18%), X2 decision-making (15%), X3 monitoring (16%), X4 predicting (51%). As a field we obtained: X5 construction process (38%), X6 occupational accident (34%), X7 risk management process (28%). The types of data investigated were: X8 construction project (5%), X9 institutional dataset (70%), X10 interview report (2%), X11 literature data (3%), X12 narrative text (6%), X13 signal (10%), X14 simulation (4%). Regarding the types of DM method, we found: X15 supervised method (58%), X16 unsupervised method (24%), X17 other method (18%). As a resource type we found: X18 process (63%), X19 environment resource (15%), X20 plant and machinery resource (22%).

3.3. Cluster Analysis

Clustering is a significant approach in DM that aims to identify groups within data sets. In real-world applications, both numeric and categorical features are often used to define the data. Clustering analysis is one of the most important approaches in DM, and it seeks to find the nature of groupings or clusters of data objects within an attribute space [8,11,16]. For an exploratory approach, we applied clustering analysis to the dataset in Appendix B. With this unsupervised ML approach, the algorithm processes input data and generates a sequence of cluster based on relational similarities with surrounding data points. The questions to answer in this DM method are: "when do we stop combining clusters?", "How do we represent clusters?". By applying Hierarchical clustering (HC) and appropriate indexes, we identified the optimal number of clusters of our data.

The “silhouette” index provided the best determination of cluster number; the highest average silhouette width indicates the optimal number of clusters [13]. The concept of silhouette width represented involves the difference between the within-cluster tightness and separation from the rest. Specifically, the silhouette width s_i for entity i∈I is defined as:

$$s_i=\frac{b_i-a_i}{\mathrm{m}\mathrm{a}\mathrm{x}(a_i,b_i)}$$

(1)

where “$a_i$” is the average distance between “i” and all other entities in the cluster to which “i” belongs and “$b_i$” is the minimum of the average distances between “i” and all entities in every other cluster. Silhouette width values are between -1 and 1. If the silhouette width value for an entity is approximately zero, it means that the entity could also be assigned to another cluster. If the silhouette width value is close to -1, it means that the entity has been incorrectly classified. If all silhouette width values are close to 1, it means that the set “i” is well clustered. The silhouette index suggests that the best aggregation of the Appendix B dataset is two clusters having obtained an index greater than 0.7 (Figure 3).

We created the item groupings through an iterative hierarchical process of aggregating pairs of “most similar” groups of methods by calculating the dissimilarity (“distance,” for triangular inequality). Thus, we obtained the dendrogram in which are represented the Euclidean distance between the elements, the similarity, and the shape of the clusters.

3.4. Principal Component Analysis (PCA)

The objective of PCA is to identify suitable Y linear transformations of the observed variables that are easily interpretable and capable of highlight and synthesise the information inherent in the initial matrix X [11,16]. This tool is particularly useful when dealing with a considerable number of variables from which one wants to extract information as possible while working with a smaller set of variables. The source data are organised in a matrix, denoted X, where the columns stand for the p observations made and the rows are the p variables considered for the phenomenon under analysis.

$$\mathrm{X}={({\mathrm{X}}_1,{\mathrm{X}}_2\dots {\mathrm{X}}_{\mathrm{p}})}^T$$

(2)

The source data matrix is synthetically represented by a multivariate random vector. Given a matrix X which holds p interrelated variables, we obtained a matrix of new data Y, consisting of p interrelated variables to each other, which turn out to be linear combinations of the former. Each principal component can be expressed as follows:

(3)

$$\begin{array}{c}\to \\ Y\end{array}=l_{ij}{X}_1+l_{ij}{X}_2+\dots l_{ip}{X}_p \mathrm{where} \mathrm{i}=1, 2\dots \mathrm{p}$$

(4)

The generic coefficient $l_{ij}$ is the weight that the variable Xj has in finding the principal component Yi (with i = 1, 2, k, p). The more the larger $l_{ij}$ is (in absolute value), the greater the weight that the values Xj (j = 1, 2, k, p) have in deciding a given principal component. The principal component Yi will be most strongly characterised by the variables Xj to which correspond the largest absolute coefficients $l_{ij}$.

Once the significance of the correlation between the different variables had been found from Formula 5, the PCA principal component analysis was applied, leaving out the outliers between the DM methods, the most important of which were analysed separately.

$$r_{xy}=Corr(y_i,x_j{)=e}_{ij}\frac{\surd \lambda i}{\sigma j}$$

(5)

The outcome of the PCA resulted in 11 components of which only the two principals of the total explained the total variance.

3.4.1. Inertia Distribution

The dataset contains 91 individuals corresponding to DM methods and 20 variables. Analysis of the graphs reveals no outlier. The inertia of the first dimensions shows if there are strong relationships between variables and suggests the number of dimensions that should be studied. The first two dimensions of analyse express 69.71% of the total dataset inertia; that means that 69.71% of the individuals (or variables) cloud total variability is explained by the plane. This percentage indicates that the first plane effectively represents the data’s variability.

The first factor is the main one: it expresses 57.36% of the variability of the data (Figure 5). In this case, the variability relating to the other components may be less significant, despite the high percentage. The first axis has a higher amount of inertia than the 0.95 quadrant of the random distributions. This observation suggests that only two axes carry factual information. Consequently, the description will stick to these axes.

The criteria for selecting dimensions in the final model are threefold: the Kaiser rule, where eigenvalues are greater than 1 (Table 4); the proportion of variance explained by the components at least equal to 60%-80% of the overall variability (Table 4); the Cattell rule, according with it the right number of components corresponds at the elbow or change of slope in the component-eigenvalue graph (Figure 5). From these observations, it should be better to also interpret the dimensions greater or equal to the second one. The above criteria allowed us to assign a “label” to each component.

3.4.2. Axes Description

Dimension 1 opposes individuals such as dt (32), knn (49), svm (81) and rf (69), to the right of the graph characterised by a strongly positive coordinate on the axis, to individuals such as MCDA C (58), characterised by a strongly negative coordinate on the axis (to the left of the graph).

Dimension 2 opposes individuals such as lstm (54), word2vec (88), nlp (63) and BIM (16), who at the top of the graph, and characterised by a low positive co-ordinate on the axis, with individuals such as ann (8), adaboost (3), who have low negative coordinate on the axis and are located at the bottom of the graph.

Figure 6. PCA. The graph of individual (DM Methods). Dim1 vs Dim2 (correlation or cos²>0.4).

Figure 6. PCA. The graph of individual (DM Methods). Dim1 vs Dim2 (correlation or cos²>0.4).

The Dim1, group 1 (dt , knn, svm and rf) is sharing high values for the variables “predicting”, “supervised”, “monitoring”, “frequency”, “institutional data”, “data project-simulation-signal”, “classifying”, “best method and “interview-literature-text” (variables are sorted from the strongest). The group 2 characterised by a negative coordinate on the axis, the individual MCDA C (58) is sharing low values for the variables “interview-literature-text”, “classifying”, “frequency”, “institutional data”, “monitoring”, “predicting”, “supervised”, “project-simulation-signal”, “best method” and “other methods (variables are sorted from the weakest). The variables “supervised” and “freq.” are highly correlated with this dimension (respective correlation of 0.94, 0.98). These variables could therefore summarize themselves the dimension 1. The Dim2, group 1 shares high values for the variables “not-supervised” and “decision-making”, while the group 2 with “monitoring”, “machinery” and “environment” (Table 5 and Table 6).

According to the correlation method-variable and axes, the x-axis (Dim1) can then be renamed "Supervised methods (dt, knn, svm and rf), applied to institutional data in classifying and making inference (predicting)". The y-axis (Dim2) can instead be renamed "Not-supervised methods (lstm, word2vec, nlp, BIM) applied to project, simulation signal, interviews, literature, or textual data in making decisions and classifying.

3.5. Meta-Analysis

The data from the complete collection of studies selected according to the PRISMA methodology, aggregated according to the classes defined in Table 1, allowed us to derive a single conclusive result that answered our research question. Through meta-analysis, we assessed whether supervised methods were more effective than not-supervised ones, for the various classes. The forest plot summarises the results of the meta-analysis, which include the OR with its CIs, the sample size weight, the heterogeneity of the data and a quantitative, whole-data assessment of the effectiveness of the treatment with supervised methods (Figure 7).

The heterogeneity is null (the sets under study are compatible). The analysis of the groups shows that the CIs intercept the "no effect" line and lose significance when taken individually, however, they are consistently overlapping and like each other. The Figure 7 indicate a general positive trend towards data treatment with supervised methods (on the left from “no-effect” line) which can be summarised by the OR = 0.71 and a CI (0.53-0.96).

4. Discussion and Future Directions

Studies focusing on DM techniques applied to the construction industry are relatively recent, dating back to 2014 at the latest and therefore, the review date from 2014 to September 2023. The number of articles on this sector has increased from 1 in 2014 to 23 in 2023. Similarly, the evolution of the total number of applied DM techniques increased from 5 between 2014 and 2016 to approximately 60 in 2023 (data not yet completed at the time of the survey).

In construction process field, 20 out of 61 observations were made regarding the construction of buildings, dams, roads, and tunnels. Within this field, the 60 out of 202 observations covered topics such as construction delays [22], crane, drilling and excavation tasks [14,18,24,39,48,70,74]; geological conditions [54], scaffolding collapse [68]; transport delays [31]; tunnelling [28,36,37,41,43,55,67]; workers and machinery location [34,40]. According to Erzaij et al. [22], project suspensions are among the most persistent tasks facing the construction sector, due to the difficulty of the industry and the essential interdependence between the bases of delay risk. The influence of delays can lead to increased time, costs, disputes, litigation, and overall rejection. The study aims to develop a data prediction tool to examine and learn the sources of delay based on previous data from construction projects, using decision trees and Bayesian naïve classification algorithms. Kumari et al. [34], investigated a machine learning architecture for excavator position detection by Global Positioning System (GPS), which can guarantee an excavator and driver position remarkably close to the real one. Wang J. et al. [67], used the principal component analysis (PCA) approach to select input factors for the prediction of tunnel boring machine (TBM) performance, particularly the travel speed. Liu et al [42], developed a model capable of predicting tunnel boring machine disc replacements based on a binary classification algorithm of Gaussian kernel support vector type cutting performance. After being trained over a period of historical data, the proposed model can predict whether cutter disc replacement is necessary, thus reducing the time required for periodic inspections. Lin et al. [40], investigated the feasibility of a real-time location service system using the Wi-Fi fingerprinting algorithm for safety risk assessment of tunnel workers. A location algorithm based on signal strength (RSS) and an artificial neural network (ann) was used for location analysis and risk assessment. Wei [68], developed wind speed prediction models based on various deep learning and machine learning techniques, in particular deep neural networks, neural networks with short-term memory, support vector regressions, random forest, and k-nearest neighbours. Subsequently, the author analysed the wind force on the scaffold and assessed the probability of the scaffold collapsing under the action of the wind.

In Occupational accident field, 16 out of 61 papers dealt with data on accidents and injuries at work from 2014 to 2023. In the class “occupational accidents,” the 89 out of 202 submissions covered the following topics: reporting of accidents [3,26,27,29,32,35,43,45,49,59,60,61,62,75,76] and days away from work. On this topic, Yelda et al. analysed textual narratives to predict injury outcome and days off work in a mining operation. For this purpose, they used decision trees, random forest and ANN and the performance of these models was compared with that of logistic regression [71]. Lee et al. [35] proposed an optimised data preprocessing method to minimise variables and main elements in diverse and complex work accident data and built an ML prediction model to achieve this. Specifically, they analysed the correlations using a flood flow diagram and applied clustering and principal component analysis (PCA) to analyse the relationships between the main variables and to be able to draw broader conclusions. However, accidents are unevenly recorded in narrative form. To date, unfortunately, there is little research on text mining, natural language processing (NLP) and deep learning (DL) techniques for analysing construction accident narratives [7]. Construction accident reports hold a wealth of empirical knowledge that could be used to better understand, predict, and prevent the occurrence of accidents in the construction sector. Large construction companies and federal agencies, such as the Occupational Safety and Health Administration (OSHA), hold these reports in the form of huge digital databases [76]. Zhang J. et al. [76], utilised accident narrative data obtained from the official OSHA website, presenting a new unified architecture having a bi-directional short-term memory model (BiLSTM) and a convolutional layer for classification of construction accident causes. Tixier et al. [60], and Zhan F. et al. [75] proved how the study of safety attributes and outcomes can be automatically and accurately processed from unstructured accident reports using natural language processing (NLP).

In risk management field 25 out of 61 papers dealt with data on the “risk management process”. Into this class the 54 observations out of 202 concerned the following topics: awkward working postures [7,19]; compliance with Health and Safety standards [2,4,47]; risk assessment [21,51,53,63,66,77,79]; safe climate [44]; slope instability [10,17]; teaching-training tasks [5,6,78]; unsafe behaviours [25,72]; worker fatigue-heat stress [38,69,73]; site image [1,46]. Antwi-Afari et al. [7] used deep learning networks to automatically extract relevant features with spatial-temporal dependence acquired by a wearable insole pressure system. The aim was to use deep learning-based networks and sensor data from wearable insoles to automatically recognise and classify types of awkward working postures for construction workers. So, they adopted recurrent neural networks (RNNs), deep learning models to train time series of plantar pressure data acquired from a wearable insole pressure sensor. Wang F. et al. [66] provided a strategic view of the relationships between different organisational objectives and technical risks that may arise during the construction of a tunnel. They created a systems-based model integrating Systems Dynamics (SD), Bayesian Belief Networks (BBN) and Smooth Relevance Vector Machines (sRVMs) called Organisational Risk Dynamics Observer (ORDO). The model was applied to an urban metro project built in Wuhan, China, and was used to provide guidance on effective accident prevention strategies. Mostofi et al. [45], explored the predictive ability of a multilayer GCN algorithm that learns the connection between construction accidents and project types, believing that richer information from existing safety and construction accident datasets by project type would provide better learning for the predictive model adopted. In addition, it would have supplied more information to predict the severity of accident consequences. The authors proved the effectiveness of the network representation of construction accidents in improving the learning capability of the ML model by using a feedforward reference network (FFN) algorithm with parameters like those used in the GCN algorithm to predict severity outcomes. The use of prefabricated is attracting increasing interest in the construction industry due to sustainability aspects, product quality, high production efficiency and cost-effectiveness. Dealing with this topic, Zhu & Liu [79] developed a prediction and risk assessment model related to the supply chain management of precast buildings. The BP neural network can be used to predict the risk of the prefabrication supply chain.

Soil instability and landslides are a major problem in the construction sector that can lead to safety risks for workers and the public, but also to considerable economic damage due to work stoppages. In this regard, Bay et al. [10] evaluated 102 cases of slopes with arc-shaped failure modes using eight machine learning regression methods. The slope safety factor prediction models were set up by performing cross-validation and hyper-parameter adjustment of the model. Furthermore, based on objective weighting and TOPSIS methods, was developed a model to evaluate the performance of the machine learning model and find the best FOS prediction model. Sadeghi et al. [53] developed an Ensemble Predictive Safety Risk Assessment Model (EPSRAM) to assess the health and safety risks of workers on construction sites based on the integration of neural networks and fuzzy inference systems. The model introduces an innovation in countries such as Malaysia, where there is continued growth in the construction industry but where there is a lack of studies on OHS assessment of workers involved in construction activities. Such circumstances may expose construction workers to the risk of developing fatigue. If workers continue to work under fatigued conditions, they are prone to the development of work-related musculoskeletal disorders (MSDs). Yu et al. [73] and Yan et al. [69], developed a combination of computer vision technology and biomechanical analysis for non-intrusive whole-body fatigue monitoring of construction workers using 3D model data from the motion capture algorithm and biomechanical analysis.

Zhao et al. [78], conducted a study on efficient and parallel DM and machine learning methods and algorithms distributed on a large scale and proposed an experiential teaching model focused on the cultivation of independent learning ability and the subjective initiative of individual learners. The article, which could have been excluded for review, was nevertheless kept as it combined the importance and technical challenges of the algorithms themselves and the context of the practical application needs of the field. It reported research on methods and algorithms for DM and machine learning, distributed on a large scale for training purposes. As an innovative teaching model, the experiential teaching model described in it, focuses among other things on cultivating individual learners’ independent learning ability and subjective initiative, which was found to effectively activate the atmosphere of the working class/environment and improve the teaching effect. It has been included as one of the articles dealing in an innovative way with risk management process, including health and safety training in the workplace. Other studies, not included in the review, report an analysis based on the effectiveness of combinations of Smart Construction Safety Technologies (SCST), potentially able to generate information useful for DM and the measurement of the effectiveness of the same technologies, single and combined [33].

Regarding the type of data used in the DM, 39 out of 61 papers dealt with institutional datasets (2016-2023), 8 used signal and video data (2014-2023), 4 narrative texts (2016-2022), 3 construction projects (2016-2023), 3 literature data (2020-2023), 3 simulations (2015-2023) and 1 out of 61 used an interview report (2023). The data used may have different characterisations, as refer to specific aspects of an occupational injury such as, for example, the body parts affected and the expected probability. Other studies focus on the observation of environmental and meteorological precursors of accidents, e.g. associated with the collapse of scaffolding [41] and slope instability [10,17]. Liu et al. analysed data from sophisticated and technologically innovative machine monitoring, capable of returning and processing geological data, faults and predicting the progress of TBM and maintenance, avoiding downtime and inspections, were analysed [43]. According to Schindler et al. [55] and Leng et al. [36], the use of satellite data has proved to be a winning strategy compared to ground surveys. Data collected by sensors was used to assess the state of effort associated with the awkward working postures taken by the worker while performing his work on the construction site [7] or the data of physical fatigue and worker's heat stress [69]. Another interesting use of data involves the construction practitioners’ interview through which it was integrated processes and occupational risk information [51]. By focusing on health and safety aspects, quality, in terms of the homogeneity and standardisation of the various sources of institutional accident data included in the review, it can be affected by the different methods of acquisition, from one institution to another and from one country to another. It can also assume that the data produced by technologies and machines used in the processes, have a higher degree of homogeneity and standardization than the former. Liu et al. underline the significance of employing innovative and efficient safety management technologies, along with new management approaches and automated methods based on artificial intelligence, to detect and eliminate risks promptly. According to the authors, these innovative technologies would mitigate any deficiencies in site management, significantly improve site safety management and eliminate risks at source [43]. An increasingly widespread orientation towards an automated management of the site or parts of it, would not only lead to an improvement in the health and safety of the processes but also a significant improvement in the quality of the data coming from the construction field. It can be assumed that soon, accident data collection techniques will not be able to do without innovative technologies capable of automatically acquiring information on near misses, accidents, and injuries in the construction sector.

Intelligent technologies can generate a range of data that pertains to both the individual (e.g. worker) and the interaction and connection between different technologies. The Internet of Things (IoT) is gradually spreading in the construction sector, thus making an important contribution to the production of new data. Robots and collaborative robots play a significant role in technological innovation and data extraction, as they can produce quality in terms of productivity, product quality, and standardization of production processes. Furthermore, these technologies have the potential to produce high-quality data, which could play a significant role in the pre-processing of data required for the use of DM techniques. The use of these technologies in construction sites is still limited due to unresolved difficulties, attributable to the high variability of environmental conditions, the need to protect the secrecy of processes and the privacy of workers. Moreover, to accompany change, workers and enterprises need vocational training and management training [30].

5. Conclusions

Cluster analysis and PCA was applied to data from articles that met the PRISMA eligibility criteria and were included in the review. The study indicates an association between types of methods used and objectives, scope, type of data and resources under investigation. This association, based on correlation, was synthesised onto a single xy-plane (Dim1 and Dim2). The results of the PCA were consistent with those of the cluster analysis. Each of the two axes was assigned a label summarising the significance of the entire review. The x-axis (Dim1) was labelled “Supervised methods (dt, knn, svm and rf) applied to institutional data for classification and inference”. The y-axis (Dim2) was labelled "Not-supervised methods (lstm, word2vec, nlp, BIM) applied to projects, simulations, signals, interviews, literature or textual data to classify and make decisions". The meta-analysis with odds ratio (OR) of 0.71 and a confidence interval (CI) of 0.53 to 0.96 provides an overall estimate of the superior effectiveness of supervised methods compared to not-supervised ones.