Eigenvector Centrality Measure Research Articles

Changing roles in global health governance following COVID-19.

The Global Health Governance (GHG) response to the COVID-19 pandemic has been criticized, particularly regarding vaccine management, and changes in the roles of GHG actors have been recommended. To investigate the perception of experts regarding changes in the roles of different GHG actors following the COVID-19 pandemic. This study used a 3-round Delphi survey to collect data from 30 global health experts between May and December 2022. The GHG roles investigated were stewardship, production of guidelines and policies, promotion of solidarity and collaboration, and management of global health challenges. Social network analysis was performed and collected data was converted into a 1-mode network. Degree centrality and Eigenvector centrality were calculated using the UCINET 6.757 modelling programme. There were variations between the current and future roles in degree centrality and eigenvector centrality for the 19 GHG actors in each of the 4 functions investigated. For stewardship, WHO, governments and the World Bank had the highest degree centrality and eigenvector centrality during both the current and future periods. In terms of production of guidelines and policies, WHO maintained the highest current and future eigenvector centralities, while research agencies, UNICEF and Gavi upheld their current eigenvector centrality measure. For the promotion of solidarity and collaboration, WHO had the highest centrality measures, followed by UNICEF, governments and Gavi. Regarding the function "management of global health challenges", WHO lost its position to UNICEF as the most central, while UNDP, FHI 360 and research agencies were predicted to have a more central role in the future. The findings position WHO as the current and future top actor in stewardship, production of guidelines and policies, and promoting solidarity and collaboration, and UNICEF as the upcoming most central actor in managing global health challenges. Governments were major actors in all GHG functions except for managing global health challenges. Funding actors were central in all GHG functions, indicating finance as an important factor in obtaining a central role in GHG. Research organizations received a high centrality rating, indicating their importance in GHG.

Using Centrality Based Topology Control for FANET Survivability Against Jamming

Flying Ad Hoc Networks (FANETs) provide connectivity in extreme environments such as battle zones and disaster areas. Since these networks operate without infrastructure, they must calculate routes on the fly. Unfortunately, as each e2e link selfishly chooses the shortest path for its flow, they also tend to overwhelm geographically central nodes, making these nodes the bottleneck for all flows. On top of this, these networks are quite sensitive to jamming attacks, which can shut down the network’s operation entirely. In this paper, we extend our previous work CentAir with an additional operation state that enables rapid communication by reducing interframe spaces of nodes according to their eigenvector centrality measure. Following this, we propose an intelligent heuristic to enable hybrid operation among these states. This hybrid operation state improves e2e latency by modifying interframe spacing regarding the node’s eigenvector centrality. Our exhaustive simulation results show that Hybrid CentAir achieves an 18.1 percent more throughput, a 21.1 percent lower latency while being 17.8 percent more energy efficient compared to multi-channel OLSR. Additionally, our protocol shows an 11.5 percent better packet delivery ratio against randomly placed deceptive jammers that continuously transmit false packets in the area.

A multilayer network framework for soccer analysis

In this paper, we define a novel methodology for analyzing soccer matches and teams using spatial multilayer networks. Departing from a segmentation of the pitch into h×v regions, we create 2-layer networks that capture the exchange of ball possessions between teams throughout a match. To assess the significance of each node, we employed eigenvector centrality measures within the constructed multilayer network. Furthermore, we introduce three additional metrics, namely the leakage, recovery and switching factor, which quantify the possession transitions between layers. Finally, we apply our methodology to analyze the performance of Spanish soccer teams over an entire season, using the aforementioned multilayer parameters, and discuss the relation with the playing style and ranking of soccer teams.

Analysis of garbage collection network of Colombo district using centrality measures

Waste management is a common problem faced by all developing countries. Colombo city faces the biggest garbage problem than other cities in Sri Lanka. Even through many studies were carried out for waste management problem using different approaches, there were very few research findings were available using graph theoretical approach. In this research, applications of graph theory in garbage collection procedure are depicted. The study mainly focuses on analyzing the garbage collection procedure of Colombo municipal council area in western province through constructing garbage collection network and using centrality measures. Centrality measures are used to compute the importance of any node in a network. Colombo municipal area divides into 6 main administrative districts, and each of these is divided into municipal wards with several junctions and streets. Garbage collection network was initially constructed by assigning a node by a location in google map, and an edge by a street or a path between two locations. Constructed network is an undirected unweighted graph and betweenness, closeness, degree, and eigenvector centrality measures are used to find central locations of the network. By identifying central locations, some machines or recycling trucks can be placed in that central places to deposit the waste. Next, a weighted graph was constructed by taking the weights of an edge as a fraction of weight of collected garbage between two locations. Collected garbage weights and betweenness and degree centrality values for weighted graph are used to identify the shortest paths between central nodes in each municipal ward. Garbage collection trucks can be followed this shortest path in order to reduce their fuel cost and collection time.

Poor Treatment Response in Panic Disorder Patients with Suicide Attempts and Their Symptom Network Characteristics

Background. Panic disorder (PD) is associated with suicidality. Depression has been suggested as a link between PD and suicide; however, this remains controversial. Comprehensive research on the history of suicide attempt (SA) in patients with PD is scarce. We investigated the clinical characteristics of SA in patients with PD using PD-related assessments and network approaches. Methods. A total of 1151 participants were enrolled, including 755 patients with PD (97 with SA (PD+SA) and 658 without SA (PD-SA)) and 396 healthy controls. The Scale for Suicide Ideation and Panic Disorder Severity, Anxiety Sensitivity Inventory, and other PD-related measures were also administered. We compared symptom severity and analyzed the pharmacological treatment response in patients with PD with and without SA. Network analysis was used to estimate the centrality, stability, and network structures of the nodes. Results. Our results revealed that the scores for panic and depressive symptoms, pathological worry, anxiety sensitivity, and the frequency of early trauma were significantly higher in the PD+SA group than in the PD-SA group. Multiple linear regression analysis revealed that short- and long-term pharmacological treatment responses were significantly poorer in the PD+SA group. Network analysis showed that fear of cognitive dyscontrol (FCD), as a cognitive aspect of anxiety sensitivity, was the central symptom through strength, expected influence (one and two steps), randomized shortest path betweenness, and eigenvector centrality measures in the PD+SA group. In contrast, depression was the central symptom of patients with PD-SA. Conclusion. Our study suggests that a history of SA could be associated with high panic-symptom severity and poor pharmacological treatment response in patients with PD and that FCD is the most central symptom in the PD+SA network. Central symptoms, such as cognitive aspects of AS in patients with PD+SA, may be clinically effective as potential targets for intervention in patients with PD at risk of or suffering from suicidality.

Eigenvector Centrality Characterization on fMRI Data: Gender and Node Differences in Normal and ASD Subjects : Author name.

With the budding interests of structural and functional network characteristics as potential parameters for abnormal brains, an essential and thus simpler representation and evaluations have become necessary. Eigenvector centrality measure of functional magnetic resonance imaging (fMRI) offer region wise network representations through fMRI diagnostic maps. The article investigates the suitability of network node centrality values to discriminate ASD subject groups compared to typically developing controls following a boxplot formalism and a classification and regression tree model. Region wise differences between normal and ASD subjects primarily belong to the frontoparietal, limbic, ventral attention, default mode and visual networks. The reduced number of regions-of-interests (ROI) clearly suggests the benefit of automated supervised machine learning algorithm over the manual classification method.

Suicide Attempts in Panic Disorder: Clinical Effects on Treatment Response and Link to Fear of Cognitive Dyscontrol

IntroductionPanic disorders (PD) are associated with suicidality. The link between PD and suicide has been suggested to be depression; however, this remains controversial. Comprehensive research on the history of suicide attempts (SAs) in patients with PD is scarce.ObjectivesThis study investigated the characteristics and pharmacological short- and long-term treatment responses of patients with PD, with or without SAs. Moreover, a network analysis was conducted to investigate the central symptoms and their connection to suicidality among SA-related variables with and without SAs.MethodsWe investigated the characteristics of SAs in patients with PD using PD-related scales, magnetic resonance imaging, and network approaches. A total of 1151 subjects were enrolled, including 755 patients with PD (97 with SA and 658 without SA) and 396 healthy controls. Suicide and PD-related scales were also administered.ResultsOur results revealed that the scores of all symptom severities were significantly higher in the PD+SA group than in the PD-SA group. Multiple linear regression analysis revealed that short- and long-term pharmacological treatment responses were significantly poor in the PD+SA group. Network analysis showed that fear of cognitive dyscontrol (FCD) was the strongest central symptom among strength, expected influence (1 and 2 step), randomized shortest path betweenness, and eigenvector centrality measures in PD+SA, whereas depression was the central symptom in PD-SA.Table 1.Results of multiple regression analysis to predict treatment response for patients with panic disorder.Treatment responseat 8 weeks(n = 450)[R2 =0.19]Treatment responseat 6 month(n = 379)[R2 =0.20]Treatment responseat 1 year(n = 329)[R2 =0.22]Βp-valueβp-valueβp-valueGender0.100.150.140.090.080.39Age-0.060.440.040.620.010.94Baseline PDSS total score0.46<0.001**0.48<0.001**0.42<0.001**Baseline BDI-II total score0.050.650.090.470.170.19Baseline PSWQ total score-0.070.47-0.110.29-0.180.09Baseline ASI-R total score-0.140.19-0.190.10-0.190.11Baseline ETISR-SF total score0.070.340.040.600.110.24A history of the suicide attempt-0.190.01*-0.200.02*-0.280.002*Note: Model p-values <0.001.*p < 0.05.**p < 0.001.Abbreviations: PD, panic disorder; SA, suicide attempt; PDSS, Panic Disorder Severity Scale; BDI-II, Beck Depression Inventory-II; PSWQ, Penn State Worry Questionnaire; ASI-R, Anxiety Sensitivity Inventory-Revised; ETISR-SF, The Early Trauma Inventory Self Report-Short Form.Image:Image 2:Image 3:ConclusionsOur results suggest that SA history could be associated with high symptom severity and poor pharmacological treatment response in patients with PD and that FCD is the central symptom in the PD+SA network.Disclosure of InterestNone Declared

Security and Makespan Trade-Off Strategy in Fog-Enabled IoT Networks

Fog computing relieves the Internet of Things (IoT) from a heavy burden of workload computation, consequently minimizes workload makespan, and reduces the considerable energy consumption of IoT devices. However, the fog-enabled IoT poses a risk to data security due to the intrinsic open and distributed structure of the fog. Hence, it is imperative to search for an optimal scheduling strategy on fog nodes to minimize workload makespan, while guaranteeing high data security. In this study, we investigate the trade-off between security and makespan and propose a multiobjective optimization model to optimize them simultaneously. To achieve this goal, we design an evolutionary algorithm together with two effective measures to handle large-scale partitionable workloads: a conflict-free multi-installment scheduling (ConfMIS) strategy as well as a security criterion based on Eigenvector centrality measure. Experimental results show that the proposed algorithm can generate a set of widely spread and uniformly distributed solutions in the decision space. Among these representative solutions, system designers/decision-makers can choose their favorite one based on specific IoT application demands.

Continuous-time quantum walk based centrality testing on weighted graphs

Centrality measure is an essential tool in network analysis and widely used in the domain of computer science, biology and sociology. Taking advantage of the speedup offered by quantum computation, various quantum centrality measures have been proposed. However, few work of quantum centrality involves weighted graphs, while the weight of edges should be considered in certain real-world networks. In this work, we extend the centrality measure based on continuous-time quantum walk to weighted graphs. We testify the feasibility and reliability of this quantum centrality using an ensemble of 41,675 graphs with various topologies and comparing with the eigenvector centrality measure. The average Vigna’s correlation index of all the tested graphs with all edge weights in [1, 10] is as high as 0.967, indicating the pretty good consistency of rankings by the continuous-time quantum walk centrality and the eigenvector centrality. The intuitive consistency of the top-ranked vertices given by this quantum centrality measure and classical centrality measures is also demonstrated on large-scale weighted graphs. Moreover, the range of the continuous-time quantum walk centrality values is much bigger than that of classical centralities, which exhibits better distinguishing ability to pick the important vertices from the ones with less importance. All these results show that the centrality measure based on continuous-time quantum walk still works well on weighted graphs.

The Curious Case of Centrality Measures: A Large-Scale Empirical Investigation

There has been extensive research using centrality measures in educational settings. One of the most common lines of such research has tested network centrality measures as indicators of success. The increasing interest in centrality measures has been kindled by the proliferation of learning analytics. Previous works have been dominated by single-course case studies that have yielded inconclusive results regarding the consistency and suitability of centrality measures as indicators of academic achievement. Therefore, large-scale studies are needed to overcome the multiple limitations of existing research (limited datasets, selective and reporting bias, as well as limited statistical power). This study aims to empirically test and verify the role of centrality measures as indicators of success in collaborative learning. For this purpose, we attempted to reproduce the most commonly used centrality measures in the literature in all the courses of an institution over five years of education. The study included a large dataset (n=3,277) consisting of 69 course offerings, with similar pedagogical underpinnings, using meta-analysis as a method to pool the results of different courses. Our results show that degree and eigenvector centrality measures can be a consistent indicator of performance in collaborative settings. Betweenness and closeness centralities yielded uncertain predictive intervals and were less likely to replicate. Our results have shown moderate levels of heterogeneity, indicating some diversity of the results comparable to single laboratory replication studies.

How Does Travel Demand Follow the Change in Infrastructure? Multiple-Year Eigenvector Centrality Analysis

The road network is one of the most permanent elements of the physical structure of cities, and the long-term impacts should be considered for effective and efficient road network improvement. It is therefore important to catch up on how the road will be used after construction. However, we do not have much knowledge on the pattern and time lag in the change process of travel demand and supply in the real situation. To explore such changes, this study proposes to evaluate a network with eigenvector centrality (EC) measures that can evaluate the importance of nodes in a network. We believe the analysis based on topological properties by the graph theory is suitable to verify the evolution of road networks. This study analysed long-term changes over 20 years in an actual city to understand the impact of road network improvements. The EC analysis with the weights of traffic indices obtained from survey data evaluates the connectivity of road services on the supply side, and traffic concentration on the demand side.

Evaluation of the vulnerability in water distribution systems through targeted attacks

Abstract This paper presents the results of a vulnerability analysis in different water distribution system (WDS) benchmarks, performed under a framework based on a graph model that integrates topological features and hydraulic characteristics, allowing the comparison between different attack strategies and centrality measures in terms of their ability to predict the shortage of water supply. This vulnerability framework has been previously applied to electric power systems and it employs a vulnerability prediction measure to quantify the amount of damage caused in terms of the physical damage measure. Different attack strategies and centrality measures were applied to four WDS benchmarks: the New York Tunnel, the Hanoi, the Modena, and the Balerma networks. It was determined that removing the most central element and recalculating the centrality for each stage are the most damaging attack strategies. Degree, eigenvector, and Katz centrality measures presented the best performance to predict the elements that are more relevant to the network and can have a larger impact on the water supply. It was demonstrated that the vulnerability framework can be applied to the WDS in the same way it was previously applied to electric power systems. Future work will be oriented to the design of the WDS using optimization techniques to minimize the vulnerability of the network under faults that can be generated by droughts and other extreme weather conditions.

Amyloid-driven disruption of default mode network connectivity in cognitively healthy individuals.

Cortical accumulation of amyloid beta is one of the first events of Alzheimer’s disease pathophysiology, and has been suggested to follow a consistent spatiotemporal ordering, starting in the posterior cingulate cortex, precuneus and medio-orbitofrontal cortex. These regions overlap with those of the default mode network, a brain network also involved in memory functions. Aberrant default mode network functional connectivity and higher network sparsity have been reported in prodromal and clinical Alzheimer’s disease. We investigated the association between amyloid burden and default mode network connectivity in the preclinical stage of Alzheimer’s disease and its association with longitudinal memory decline. We included 173 participants, in which amyloid burden was assessed both in CSF by the amyloid beta 42/40 ratio, capturing the soluble part of amyloid pathology, and in dynamic PET scans calculating the non-displaceable binding potential in early-stage regions. The default mode network was identified with resting-state functional MRI. Then, we calculated functional connectivity in the default mode network, derived from independent component analysis, and eigenvector centrality, a graph measure recursively defining important nodes on the base of their connection with other important nodes. Memory was tested at baseline, 2- and 4-year follow-up. We demonstrated that higher amyloid burden as measured by both CSF amyloid beta 42/40 ratio and non-displaceable binding potential in the posterior cingulate cortex was associated with lower functional connectivity in the default mode network. The association between amyloid burden (CSF and non-displaceable binding potential in the posterior cingulate cortex) and aberrant default mode network connectivity was confirmed at the voxel level with both functional connectivity and eigenvector centrality measures, and it was driven by voxel clusters localized in the precuneus, cingulate, angular and left middle temporal gyri. Moreover, we demonstrated that functional connectivity in the default mode network predicts longitudinal memory decline synergistically with regional amyloid burden, as measured by non-displaceable binding potential in the posterior cingulate cortex. Taken together, these results suggest that early amyloid beta deposition is associated with aberrant default mode network connectivity in cognitively healthy individuals and that default mode network connectivity markers can be used to identify subjects at risk of memory decline.

Identifying influential links to control spreading of epidemics

In recent years, epidemics have been raging around the world, causing serious harm to human health, and controlling the spread of the epidemics has become a hot topic. The study of network partition is of great practical significance to control the spread of the epidemics. Network partition is an NP hard problem. Much work has focused on identifying important nodes in complex networks to achieve network partition. Similarly, understanding the importance of links in spreading dynamics in a network can provide ways to hinder or slow down ongoing dynamical phenomena like the spreading of epidemic or the diffusion of information. This paper studies the algorithms to identify influential links in complex networks. Deleting some influential links can divide the network into some components so as to prevent effectively the spreading processes in epidemics. Our algorithms can successfully identify influential links at low computational cost by utilizing the eigenvector centrality measures. The simulated results in some real-world social networks and synthetic networks show that our algorithms are effective. The number of infected individuals can be remarkable reduced by removing influential links.

Rheostat functional outcomes occur when substitutions are introduced at nonconserved positions that diverge with speciation.

When amino acids vary during evolution, the outcome can be functionally neutral or biologically-important. We previously found that substituting a subset of nonconserved positions, "rheostat" positions, can have surprising effects on protein function. Since changes at rheostat positions can facilitate functional evolution or cause disease, more examples are needed to understand their unique biophysical characteristics. Here, we explored whether "phylogenetic" patterns of change in multiple sequence alignments (such as positions with subfamily specific conservation) predict the locations of functional rheostat positions. To that end, we experimentally tested eight phylogenetic positions in human liver pyruvate kinase (hLPYK), using 10-15 substitutions per position and biochemical assays that yielded five functional parameters. Five positions were strongly rheostatic and three were non-neutral. To test the corollary that positions with low phylogenetic scores were not rheostat positions, we combined these phylogenetic positions with previously-identified hLPYK rheostat, "toggle" (most substitution abolished function), and "neutral" (all substitutions were like wild-type) positions. Despite representing 428 variants, this set of 33 positions was poorly statistically powered. Thus, we turned to the in vivo phenotypic dataset for E. coli lactose repressor protein (LacI), which comprised 12-13 substitutions at 329 positions and could be used to identify rheostat, toggle, and neutral positions. Combined hLPYK and LacI results show that positions with strong phylogenetic patterns of change are more likely to exhibit rheostat substitution outcomes than neutral or toggle outcomes. Furthermore, phylogenetic patterns were more successful at identifying rheostat positions than were co-evolutionary or eigenvector centrality measures of evolutionary change.

Faster heuristics for graph burning.

Graph burning is a process of information spreading through the network by an agent in discrete steps. The problem is to find an optimal sequence of nodes that have to be given information so that the network is covered in least number of steps. Graph burning problem is NP-Hard for which two approximation algorithms and a few heuristics have been proposed in the literature. In this work, we propose three heuristics, namely, Backbone Based Greedy Heuristic (BBGH), Improved Cutting Corners Heuristic (ICCH), and Component Based Recursive Heuristic (CBRH). These are mainly based on Eigenvector centrality measure. BBGH finds a backbone of the network and picks vertex to be burned greedily from the vertices of the backbone. ICCH is a shortest path based heuristic and picks vertex to burn greedily from best central nodes. The burning number problem on disconnected graphs is harder than on the connected graphs. For example, burning number problem is easy on a path where as it is NP-Hard on disjoint paths. In practice, large networks are generally disconnected and moreover even if the input graph is connected, during the burning process the graph among the unburned vertices may be disconnected. For disconnected graphs, ordering the components is crucial. Our CBRH works well on disconnected graphs as it prioritizes the components. All the heuristics have been implemented and tested on several bench-mark networks including large networks of size more than 50K nodes. The experimentation also includes comparison to the approximation algorithms. The advantages of our algorithms are that they are much simpler to implement and also several orders faster than the heuristics proposed in the literature.

Hashtag activism and message frames: social network analysis of Instagram during the COVID-19 pandemic outbreak in Indonesia

During the Coronavirus pandemic, social media played an essential role in disseminating information online by utilizing various message frames and hashtags. Instagram is a popular social media worldwide that are used to spread various information. This study applied the social network analysis (SNA) approach as a theoretical framework and explored how the relationship between users on the Instagram platform related to Coronavirus outbreaks. Moreover, we also investigate how the relationship between users and their hashtags. SNA techniques were used for visualizing network models using an undirected graph, measuring network attributes, and centrality measures to find the most influential users in the network. A total of 10,403 posts based on #wabahcorona on Instagram from February 28, 2020, to May 18, 2020, were analyzed. Based on the analysis results, hashtags play an important role in this topic. Degree centrality measure as connectivity number shows that only two user accounts can make it into the list’s top ten. When using the Eigenvector centrality measure, there are no users in the top ten. The modularity measure detects 122 distinct communities that show the dense connection of nodes in the networks. Betweenness centrality measure shows that there are six related hashtags to COVID-19 pandemic out of ten hashtags. They are #wabahcorona, #covid19, #corona, #viruscorona, #dirumahaja, and #coronavirus and the four are Islamic related hashtags. The Islamic related hashtags are caused by the date of Ramadhan month.

Finding the resistance distance and eigenvector centrality from the network’s eigenvalues

There are different measures to classify a network’s data set that, depending on the problem, have different success rates. For example, the resistance distance and eigenvector centrality measures have been successful in revealing ecological pathways and differentiating between biomedical images of patients with Alzheimer’s disease, respectively. The resistance distance measures an effective distance between two nodes of a network taking into account all possible shortest paths between them and the eigenvector centrality measures the relative importance of each node in a network. However, both measures require knowing the network’s eigenvalues and eigenvectors. Here, we show that we can closely approximate [find exactly] the resistance distance [eigenvector centrality] of a network only using its eigenvalue spectra, where we illustrate this by experimenting on resistor circuits, real neural networks (weighted and unweighted), and paradigmatic network models — scale-free, random, and small-world networks. Our results are supported by analytical derivations, which are based on the eigenvector–eigenvalue identity. Since the identity is unrestricted to the resistance distance or eigenvector centrality measures, it can be applied to most problems requiring the calculation of eigenvectors.

Influential nodes in planar trusses and meshes using centrality measures

Centrality means the quality or state of staying in the central. Centrality measures in graph theory are used to find the most influential nodes in a graph. Identifying the influential nodes in the truss structures is important to understand the responsible node for design and failure of the structures. Targeting the influential nodes in a structure such as a truss or a mesh is of immense help in optimization of support during actual construction. In this paper, the uses and applications of a planar truss, the elemental clique and skeleton graphs of a Finite Element Mesh (FEM) are discussed. Various sizes of these structures are simulated digitally using python programming language and are analyzed to derive a general formula in order to find the most influential nodes on the basis of degree, closeness, betweenness and eigenvector centrality measures. The findings of this study can be used to provide decision – making support for the analysis, design and optimization of the truss structures.

Discovering latent themes in traffic fatal crash narratives using text mining analytics and network topology

The proliferation of digital textual archives in the transportation safety domain makes it imperative for the inventions of efficient ways of extracting information from the textual data sources. The present study aims at utilizing crash narratives complemented by crash metadata to discern the prevalence and co-occurrence of themes that contribute to crash incidents. Ten years (2009–2018) of Michigan traffic fatal crash narratives were used as a case study. The structural topic modeling (STM) and network topology analysis were used to generate and examine the prevalence and interaction of themes from the crash narratives that were mainly categorized into pre-crash events, crash locations and involved parties in the traffic crashes. The main advantage of the STM over the other topic modeling approaches is that it allows the researchers to discover themes from documents and estimate how the topic relates to the document metadata. Topics with the highest prevalence for the angle, head-on, rear-end, sideswipe and single motor vehicle crashes were crash at stop-sign, crossing the centerline, unable to stop, lane change maneuver and run-off-road crash, respectively. Eigenvector centrality measure in network topology showed that event-related topics were consistently central in articulating the crash occurrence. The centrality and association between topics varied across crash types. The efficacy of generated topics in classifying crashes by type was tested using a machine learning algorithm, Random Forest. The classification accuracy in the held-out sample ranged between 89.3 % for sideswipe crashes to 99.2 % for single motor vehicle crashes. High classification accuracy suggests that automation of crash typing and consistency checks can be accomplished effectively by using extracted latent themes from the crash narratives.