Clustering Modeling Research Articles

Comparing galaxy clustering in Horizon-AGN simulated light-cone mocks and VIDEO observations

ABSTRACT Hydrodynamical cosmological simulations have recently made great advances in reproducing galaxy mass assembly over cosmic time – as often quantified from the comparison of their predicted stellar mass functions to observed stellar mass functions from data. In this paper, we compare the clustering of galaxies from the hydrodynamical cosmological simulated light-cone Horizon-AGN to clustering measurements from the VIDEO survey observations. Using mocks built from a VIDEO-like photometry, we first explore the bias introduced into clustering measurements by using stellar masses and redshifts derived from spectral energy distribution fitting, rather than the intrinsic values. The propagation of redshift and mass statistical and systematic uncertainties in the clustering measurements causes us to underestimate the clustering amplitude. We then find that clustering and halo occupation distribution (HOD) modelling results are qualitatively similar in Horizon-AGN and VIDEO. However, at low stellar masses, Horizon-AGN underestimates the observed clustering by up to a factor of ∼3, reflecting the known excess stellar mass to halo mass ratio for Horizon-AGN low-mass haloes, already discussed in previous works. This reinforces the need for stronger regulation of star formation in low-mass haloes in the simulation. Finally, the comparison of the stellar mass to halo mass ratio in the simulated catalogue, inferred from angular clustering, to that directly measured from the simulation validates HOD modelling of clustering as a probe of the galaxy–halo connection.

A Framework Based on K-Means Clustering and Topic Modeling for Analyzing Unstructured Manufacturing Capability Data

Abstract The natural language descriptions of the capabilities of manufacturing companies can be found in multiple locations including company websites, legacy system databases, and ad hoc documents and spreadsheets. To unlock the value of unstructured capability data and learn from it, there is a need for developing advanced quantitative methods supported by machine learning and natural language processing techniques. This research proposes a hybrid unsupervised learning methodology using K-means clustering and topic modeling techniques in order to build clusters of suppliers based on their capabilities, automatically infer topics from the created clusters, and discover nontrivial patterns in manufacturing capability corpora. The capability data is extracted either directly from the website of manufacturing firms or from their profiles in e-sourcing portals and directories. Feature extraction and dimensionality reduction process in this work are supported by N-gram extraction and latent semantic analysis (LSA) methods. The proposed clustering method is validated experimentally based on a dataset composed of 150 capability descriptions collected from web-based sourcing directories such as the Thomas Net directory for manufacturing companies. The results of the experiment show that the proposed method creates supplier cluster with high accuracy. Two example applications of the proposed framework, related to supplier similarity measurement and automated thesaurus creation, are introduced in this paper.

Novel trajectory clustering method based on distance dependent Chinese restaurant process

Trajectory clustering and path modelling are two core tasks in intelligent transport systems with a wide range of applications, from modeling drivers’ behavior to traffic monitoring of road intersections. Traditional trajectory analysis considers them as separate tasks, where the system first clusters the trajectories into a known number of clusters and then the path taken in each cluster is modelled. However, such a hierarchy does not allow the knowledge of the path model to be used to improve the performance of trajectory clustering. Based on the distance dependent Chinese restaurant process (DDCRP), a trajectory analysis system that simultaneously performs trajectory clustering and path modelling was proposed. Unlike most traditional approaches where the number of clusters should be known, the proposed method decides the number of clusters automatically. The proposed algorithm was tested on two publicly available trajectory datasets, and the experimental results recorded better performance and considerable improvement in both datasets for the task of trajectory clustering compared to traditional approaches. The study proved that the proposed method is an appropriate candidate to be used for trajectory clustering and path modelling.

From ""A"" to ""THE"" Behavioral Theory of the Firm: A Historical Review based on Text Mining

This study reviews 564 articles in the past 55 years in the field of the Behavioral Theory of the Firm (hereafter BToF) and reveals the evolutionary process from “A” to “THE” BToF with text mining techniques. Our findings suggest that the development of BToF experienced four phases and themes change in different phases, demonstrating an overall picture of how “A” BToF evolves to “THE” BToF. The finding of co-word network analysis and text clustering and topic modeling approaches further reveal the changing structure of BToF. 14 topics are identified and compared, revealing that performance feedback and aspirations are more likely to be the future research focus. Comparing the average topic weight of the top-cited articles with all articles, we uncover that “A” BToF has a greater academic impact than “THE” BToF which reminds us to go back to “A” BToF and revisit some neglected territories in previous studies. What’s more, BToF has a broader influence on many other theories, such as evolutionary economic and organizational learning theory, while it lacks its own academic identity. Additional bibliometric methods, like author partnership network, are conducted to supplement the results of the text mining techniques.

Text Mining and Subject Analysis for Fiction; or, Using Machine Learning and Information Extraction to Assign Subject Headings to Dime Novels

This article describes multiple experiments in text mining at Northern Illinois University that were undertaken to improve the efficiency and accuracy of cataloging. It focuses narrowly on subject analysis of dime novels, a format of inexpensive fiction that was popular in the United States between 1860 and 1915. NIU holds more than 55,000 dime novels in its collections, which it is in the process of comprehensively digitizing. Classification, keyword extraction, named-entity recognition, clustering, and topic modeling are discussed as means of assigning subject headings to improve their discoverability by researchers and to increase the productivity of digitization workflows.

An Assessment of Groundwater Contamination Risk with Radon Based on Clustering and Structural Models

There is currently some controversy in the scientific community regarding the efficiency of the water–rock interaction process in the contamination of radon in groundwater. In this study, some difficulties were found in the sampling phase. Many of the water collection points are used for human consumption. As such, some municipalities did not want to collaborate. When this natural contaminant is undetectable to the human sense and may cause pulmonary neoplasms in the long term, it is difficult to obtain collaboration from the municipalities concerned. To overcome this controversy, it is important to understand that geogenic, climatic, hydrological, and topographic features may contribute to the effective transfer of radon from rocks to groundwater. In brief, this new approach combines the radon transfer from the geological substrate to the groundwater circulation through hierarchic agglomerative clustering (HAC) and partial least squares-path modeling (PLS-PM) methods. The results show that some lithologies with higher radon production may not always contribute to noticeable groundwater contamination. In this group, the high-fracturing density confirms the recharge efficiency, and the physical-chemical properties of the hydraulic environment (electric conductivity) plays the main role of radon unavailability in the water intended for human consumption. Besides, the hydraulic turnover time of the springs can be considered an excellent radiological indicator in groundwater. In the absence of an anomalous radioactive source near the surface, it means that the high-turnover time of the springs leads to a low-radon concentration in the water. Besides linking high-risk areas with a short period required to free local flow discharges, this study exposes the virtues of a new perspective of a groundwater contamination risk modeling.

Species delimitation of the Dermacentor ticks based on phylogenetic clustering and niche modeling.

Three species belonging to the genus Dermacentor (Acari: Ixodidae), D. marginatus, D. nuttalli and D. silvarum are well known as vectors for a great variety of infection pathogens. All three of them are host ticks, which are very similar in morphology characteristics, life cycle, seasonal variation and ecological conditions, making it difficult to distinguish the three species. In the present study, these three species were delimitated based on molecular data and ecological niche. The molecular analysis showed that the three species can be distinguished by COI and ITS2 sequences. We created future potential distribution maps for the three species under climate changes with MaxEnt, which highlighted the different levels of the suitable habitats for each tick species. In addition, niche comparisons among the three species in Dermacentor were conducted, and the analysis suggested that niche overlap was relatively high with D. nuttalli and D. silvarum compared to the other species pairs, which was consistent with the molecular data. Niche equivalency and similarity test confirmed that these Dermacentor species were closely related but distinct species. In conclusion, delimitation of these three species within Dermacentor was supported by molecular phylogeny and quantitative ecological space. This study will provide deep insights into the biology, ecology, and diversification processes within Dermacentor species, and for the development of effective control for ticks.

3RAD-based systematics of the transitional Nearctic-Neotropical lubber grasshopper genus Taeniopoda (Orthoptera: Romaleidae)

The genus Taeniopoda Stål (Romaleidae) is a group of Nearctic-Neotropical grasshoppers whose systematics has been largely neglected. A recent phylogenetic study based on morphology and mitochondrial and nuclear markers failed to resolve the species boundaries in this genus and showed a lack of reciprocal exclusivity between T. eques (Burmeister) and T. tamaulipensis Rehn. Here we assessed the species limits and phylogenetic relationships in Taeniopoda based on 3RAD data, and evaluated the presence of gene flow and niche overlap between the above two species using clustering and ecological niche modelling (ENM) analyses to determine their taxonomic status. We performed de novo assembly of different 3RAD data sets with distinct parameters settings to explore whether they impact the recovered relationships. Ten species were consistently delimited, with T. picticornis and T. stali regarded as conspecific and the populations of T. auricornis from Guatemala representing a separate species. We maintained the specific status of T. eques and T. tamaulipensis, though our results suggest that they represent a ring species since their genetic composition appear to change gradually following a “loop form” along their geographical distribution. The phylogenomic analyses confirmed the paraphyly of Taeniopoda with respect to Romalea and recovered three major clades. Similar to previous studies, the relationships of our examined matrices were highly congruent despite their different levels of missing data. However, the similarity threshold and minimum number of samples that must share a locus for it to be retained impact the amount of loci and missing data of the matrices. This study demonstrates the utility of 3RAD to detect gene flow and to resolve species limits and phylogenetic relationships among closely related taxa.

An evaluation of document clustering and topic modelling in two online social networks: Twitter and Reddit

Methods for document clustering and topic modelling in online social networks (OSNs) offer a means of categorising, annotating and making sense of large volumes of user generated content. Many techniques have been developed over the years, ranging from text mining and clustering methods to latent topic models and neural embedding approaches. However, many of these methods deliver poor results when applied to OSN data as such text is notoriously short and noisy, and often results are not comparable across studies. In this study we evaluate several techniques for document clustering and topic modelling on three datasets from Twitter and Reddit. We benchmark four different feature representations derived from term-frequency inverse-document-frequency (tf-idf) matrices and word embedding models combined with four clustering methods, and we include a Latent Dirichlet Allocation topic model for comparison. Several different evaluation measures are used in the literature, so we provide a discussion and recommendation for the most appropriate extrinsic measures for this task. We also demonstrate the performance of the methods over data sets with different document lengths. Our results show that clustering techniques applied to neural embedding feature representations delivered the best performance over all data sets using appropriate extrinsic evaluation measures. We also demonstrate a method for interpreting the clusters with a top-words based approach using tf-idf weights combined with embedding distance measures.

Individual play patterns stimulated by a familiar object are group-driven

This study investigates the dynamics of play behaviour within groups of four juvenile pigs and uses a novel clustering and statistical modelling approach to describe new details in how individuals play with a familiar object (toy rope). We examined complex state sequence data collected during a 30 min home pen play test, using the package TraMineR, where the states were defined as object play, locomotor/social play and no play. From behavioural observations, and based on the relative proportion of the different types of object play observed, each individual was later categorised as an initiator or joiner type of player. Initiators were found to be more solitary and to show more object play whereas joiners were more social and showed less object play. The majority of groups did not have an initiator type of player, yet on average they played more. Despite strong group and type of player effects, we identified three general individual play patterns. On a group level, our results demonstrate differences in how a period of playing develops, that playing with the object simultaneously occurs more often in groups than expected by chance and that the number of pigs playing together is stable over time.

Biological‐based habitat classification approaches promote cost‐efficient monitoring: An example using seabed assemblages

Abstract Seabed habitat maps can help facilitate the management of marine environments. A variety of approaches exist for seabed habitat classification. Most partition the environment according to physical environmental characteristics, with an assumption that resulting habitat classes are biologically meaningful. In the absence of comprehensive broad‐scale biological data, this strategy offers a logical and pragmatic way of producing habitat maps to help manage the marine environment. Across Europe, the physical based European Nature Information System (EUNIS) classification has gained wide acceptance, with maps used to classify broadscale habitats within Marine Protected Areas and to design monitoring programmes. An alternative approach to habitat classification, made possible by increasing quantities of data, is to use the biology to identify meaningful habitats. With such contrasting approaches, the question arises as to which provides the most robust and efficient basis for biological monitoring. To investigate, we compared variability in macrofaunal assemblages across different EUNIS sediment classes to those of two new habitat classification approaches developed in this study. The first of these (PHY) is based on a wide suite of physical variables known to influence the fauna. The second (BIO) uses the fauna to identify meaningful habitats. Both classifications were produced using a training dataset (9,619 grab samples) and employing k‐means clustering and Random Forest Modelling. Power analysis of test set data (4,123 samples) was used to assess the number of samples required to detect a 20% change in taxon richness and total abundance across all classes of each classification approach. Results showed that across all habitat classes, the BIO classification required 49% and 31% fewer samples to detect the change in richness and abundance than EUNIS level 4. Whilst offering some improvement on EUNIS, PHY still required many more samples than BIO. Synthesis and applications. Habitat maps based on biological data have generally lower within‐habitat variability in community metrics than those produced using physical attributes alone. As a result, biologically‐based habitat maps could offer a more cost‐effective basis for ecological monitoring.

Image offset density distribution model and recognition of hand knuckle

The accurate description of hand posture plays an important role in the man-machine interaction involved in coordinated assembly. Knuckle image extraction and recognition are of great significance to refine and enrich hand-pose information. These are based on nonparametric density kernel estimation observation sets corresponding to unilateral and bilateral excursion of the hand knuckle gray image. In this paper, sets of pixel positions belonging to the upper- and middle-density intervals are used as two types of image targets. Random clustering and random field multi-classification target modeling are used to learn and estimate the two target distributions of the image. The discriminant field classification learning method is used to fuse the two kinds of target models. A comprehensive representation of the image offset features is obtained. Finally, the knuckle image sample set is used to train the model, and the adaptive threshold is used to identify the hand knuckle image. The results show that the proposed method is feasible.

Spatial clustering of single-industry towns and a dynamic model of economic growth

This article studies the problems of economic growth and spatial development of regions with a high concentration of single-industry towns. The authors aim to identify the factors of development of single-industry towns at the microeconomic level on the basis of clustering and dynamic modeling of single-industry towns in three regions with the highest concentration — Kemerovo, Sverdlovsk, and Chelyabinsk Regions. This paper performs the clustering of single-industry towns by entropy indicators and the number of newly created enterprises, which allows distinguishing three “central” single-industry towns in each of the respective regions: Novokuznetsk, Nizhniy Tagil, and Magnitogorsk. The clustering of single-industry towns with the use of the population-normalized index of the number of newly created enterprises allows us to refer these cities to two different clusters: Novokuznetsk against Nizhniy Tagil and Magnitogorsk with different parameters dominating. The correlation analysis of aggregate revenue, fixed assets, the share of the single industry, the entropy of revenue, the number of newly created enterprises of three single-industry towns allows suggesting a dynamic regression model. The peculiarity of this model is the inclusion as a variable of the number of the newly created enterprises in Nizhniy Tagil for all the cities under consideration, as well as the inclusion of a dummy variable reflecting the year of introduction of the program for the development of single-industry towns. Using the bootstrap method for Novokuznetsk, the authors have confirmed the significance of the introduction of this variable. The results of the study have revealed both common patterns of regional development — the positive effect of reducing the share of monaurally and monocentric provision of single-industry towns in the regions for economic growth — and their differences — a contradictory effect of diversity in the central towns, as well as the assignment of Novokuznetsk and Nizhniy Tagil with Magnitogorsk from different clusters. In conclusion, the authors justify the early completion of the program of diversification of single-industry towns, designed to be ineffective in 2019.

Chemical Characteristics and Source Apportionment of PM2.5 during Winter in the Southern Part of Urumqi, China

ABSTRACT Urumqi, the administrative center of Xinjiang, suffers from severe atmospheric aerosol pollution; however, no study has comprehensively analyzed the local constituents and sources of fine particulate matter (PM2.5). The characteristics of PM2.5 in Urumqi were observed the first winter (2012–2013) after natural gas replaced coal as an energy source. Enrichment factors, backward trajectories, the potential source contribution function (PSCF) model, and positive matrix factorization (PMF) were used to identify the source area and categories. The results showed a mean concentration of 197.40 µg m–3 for the PM2.5, which significantly decreased after the conversion from coal to natural gas. Although the concentration of NO3– increased post-conversion, the SO42– and Cl– decreased by 42.54% and 32.93%, respectively. The water-soluble ions (WSIs) mainly consisted of NH4HSO4, CaSO4, MgSO4, Ca(NO3)2, Mg(NO3)2, and KCl. Elements such as Pb, Cr, and As decreased following the fuel switch. The organic carbon and elemental carbon were strongly correlated, and the mean concentration of the secondary organic carbon was 18.90 µg m–3. Pyr, Chr, BbF, BkF, IcdP, and BghiP were the most prevalent individual polycyclic aromatic hydrocarbons, and BaP exceeded health-based guidelines. The results from trajectory clustering and PSCF modeling suggested that emissions from both the city and its surroundings, as well as the valley-and-basin topography, may be responsible for the heavy PM2.5 pollution in southern Urumqi. PMF identified five primary sources: secondary formation, biomass and waste burning, vehicle emissions, crustal minerals, and industrial pollution and coal combustion.

Combining Bayesian genetic clustering and ecological niche modeling: Insights into wolf intraspecific genetic structure.

The distribution of intraspecific genetic variation and how it relates to environmental factors is of increasing interest to researchers in macroecology and biogeography. Recent studies investigated the relationships between the environment and patterns of intraspecific genetic variation across species ranges but only few rigorously tested the relation between genetic groups and their ecological niches. We quantified the relationship of genetic differentiation (F ST) and the overlap of ecological niches (as measured by n‐dimensional hypervolumes) among genetic groups resulting from spatial Bayesian genetic clustering in the wolf (Canis lupus) in the Italian peninsula. Within the Italian wolf population, four genetic clusters were detected, and these clusters showed different ecological niches. Moreover, different wolf clusters were significantly related to differences in land cover and human disturbance features. Such differences in the ecological niches of genetic clusters should be interpreted in light of neutral processes that hinder movement, dispersal, and gene flow among the genetic clusters, in order to not prematurely assume any selective or adaptive processes. In the present study, we found that both the plasticity of wolves—a habitat generalist—to cope with different environmental conditions and the occurrence of barriers that limit gene flow lead to the formation of genetic intraspecific genetic clusters and their distinct ecological niches.

Label-free tissue proteomics can classify oral squamous cell carcinoma from healthy tissue in a stage-specific manner

ObjectivesNo prognostic or predictive biomarkers for oral squamous cell carcinoma (OSCC) exist. We aimed to discover novel proteins, altered in OSCC, to be further investigated as potential biomarkers, and to improve understanding about pathways involved in OSCC. Materials and MethodsProteomic signatures of seven paired healthy and OSCC tissue samples were identified using ultra-definition quantitative mass spectrometry, then analysed and compared using Anova, principal component analysis, hierarchical clustering and OPLS-DA modelling. A selection of significant proteins that were also altered in the serum from a previous study (PMID: 28632724) were validated immunohistochemically on an independent cohort (n = 66) to confirm immunopositivity and location within tumour tissue. Ingenuity Pathways Analysis was employed to identify altered pathways. ResultsOf 829 proteins quantified, 257 were significant and 72 were able to classify healthy vs OSCC using OPLS-DA modelling. We identified 19 proteins not previously known to be upregulated in OSCC, including prosaposin and alpha-taxilin. KIAA1217 and NDRG1 were upregulated in stage IVa compared with stage I tumours. Altered pathways included calcium signalling, cellular movement, haematological system development and function, and immune cell trafficking, and involved NF-kB and MAPK networks. ConclusionsWe found a set of proteins reliably separating OSCC tumour from healthy tissue, and multiple proteins differing between stage I and stage IVa OSCC. These potential biomarkers can be studied and validated in larger cohorts.

Prioritising references for systematic reviews with RobotAnalyst: A user study.

Screening references is a time‐consuming step necessary for systematic reviews and guideline development. Previous studies have shown that human effort can be reduced by using machine learning software to prioritise large reference collections such that most of the relevant references are identified before screening is completed. We describe and evaluate RobotAnalyst, a Web‐based software system that combines text‐mining and machine learning algorithms for organising references by their content and actively prioritising them based on a relevancy classification model trained and updated throughout the process. We report an evaluation over 22 reference collections (most are related to public health topics) screened using RobotAnalyst with a total of 43 610 abstract‐level decisions. The number of references that needed to be screened to identify 95% of the abstract‐level inclusions for the evidence review was reduced on 19 of the 22 collections. Significant gains over random sampling were achieved for all reviews conducted with active prioritisation, as compared with only two of five when prioritisation was not used. RobotAnalyst's descriptive clustering and topic modelling functionalities were also evaluated by public health analysts. Descriptive clustering provided more coherent organisation than topic modelling, and the content of the clusters was apparent to the users across a varying number of clusters. This is the first large‐scale study using technology‐assisted screening to perform new reviews, and the positive results provide empirical evidence that RobotAnalyst can accelerate the identification of relevant studies. The results also highlight the issue of user complacency and the need for a stopping criterion to realise the work savings.

Discovering themes in biomedical literature using a projection-based algorithm

BackgroundThe need to organize any large document collection in a manner that facilitates human comprehension has become crucial with the increasing volume of information available. Two common approaches to provide a broad overview of the information space are document clustering and topic modeling. Clustering aims to group documents or terms into meaningful clusters. Topic modeling, on the other hand, focuses on finding coherent keywords for describing topics appearing in a set of documents. In addition, there have been efforts for clustering documents and finding keywords simultaneously.ResultsWe present an algorithm to analyze document collections that is based on a notion of a theme, defined as a dual representation based on a set of documents and key terms. In this work, a novel vector space mechanism is proposed for computing themes. Starting with a single document, the theme algorithm treats terms and documents as explicit components, and iteratively uses each representation to refine the other until the theme is detected. The method heavily relies on an optimization routine that we refer to as the projection algorithm which, under specific conditions, is guaranteed to converge to the first singular vector of a data matrix. We apply our algorithm to a collection of about sixty thousand PubMed Ⓡ documents examining the subject of Single Nucleotide Polymorphism, evaluate the results and show the effectiveness and scalability of the proposed method.ConclusionsThis study presents a contribution on theoretical and algorithmic levels, as well as demonstrates the feasibility of the method for large scale applications. The evaluation of our system on benchmark datasets demonstrates that our method compares favorably with the current state-of-the-art methods in computing clusters of documents with coherent topic terms.

Identification of representative buildings and building groups in urban datasets using a novel pre-processing, classification, clustering and predictive modelling approach

The formulation of energy policies for urban building stock frequently requires the evaluation of the energy use of large numbers of buildings. When urban energy modelling is utilised as part of this process, the identification of building groups and associated representative buildings can play a critical role. This paper outlines a novel methodology for identifying building groups and associated representative buildings in urban datasets. The methodology utilizes a combination of building classification, building clustering and predictive modelling. First, multiple urban-scale datasets are collected, and then, classification techniques and clustering algorithms are applied to identify building clusters. Next, representative buildings (medoids) in each cluster are identified. Predictive modelling is used to expand cluster membership in the case where some buildings were excluded from the previous analysis. A number of different clustering algorithms are assessed, including K-means and hierarchical (agglomerative and divisive) and partitioning around medoids. The methodology is applied to a large dataset of mixed-use buildings in the city of Geneva, Switzerland. The results, assessed by nine validation indices, indicate the capacity of the decision support framework to identify clusters and associated representative buildings. Furthermore, post-application of predictive modelling, using a random forest approach, facilitates the incorporation of a larger portion of the building stock within the established clusters with an overall average classification accuracy of 89%. A total of 67 representative buildings were identified in the urban dataset, which consisted of 13614 mixed-use buildings in the city of Geneva.

Optimal delivery routing with wider drone-delivery areas along a shorter truck-route

While convergent technology has been booming recently, drones are being applied in various fields of industry and are expected to be used as a commercial delivery method. In a delivery system, routing becomes one of the major issues, and several studies have attempted to solve drone-based routing problems. In this study, we focus on finding an effective delivery route for trucks carrying drones. To put it concretely, we propose a new approach on a nonlinear programming model to find shift-weights that move the centers of clusters to make for wider drone-delivery areas along shorter truck-route after initial K-means clustering and TSP (Traveling Salesman Problem) modeling. In order to verify the effectiveness of the proposed model with shift-weights, we compare it with two other delivery route approaches. One is a route without shift-weights after K-means clustering and TSP modeling, and the other is a route by TSP for all delivery locations without K-means clustering. Through experimental results of paired t-tests on randomly generated delivery locations, we show that our proposed model is more effective than the other two models.