Combination Of Cluster Analysis Research Articles

Clustering approach to urban rainfall time series prediction with support vector regression model

ABSTRACT The aim of rainfall prediction in urban areas is to provide information about precipitation, which may cause flooding resulting from the insufficient capacity of sewage systems. The research combined cluster analysis and support vector regression (CSVR) to create a simpler, and as effective as hybrid models, method predicting daily rainfall in Wroclaw, Poland. Results demonstrate that the clustering approach while modelling is reasonable, improves the quality of prediction and minimises error values. The best SVR performance, RMSE = 2.492 mm and r2 = 0.830 in the testing subset, was obtained for the Ward clustering method. Nevertheless, comparisons with MLP prediction, combined with k-means clustering, proved to be slightly more accurate and led to the creation of a model with RMSE = 1.974 and r2 = 0.899. The presented approach might constitute an alternative method to be used for creating rainfall patterns in stormwater discharge or management systems leading to an increase in flash flood safety in an urban catchment.

Droughts’ projections in homogeneous climatic regions using Standardized Precipitation Index in Pakistan

This study is conducted to develop homogenous climatic regions (HCR) based on Standardized Precipitation Index (SPI) by combining cluster analysis and L-moment approach. Initially, cluster analysis was performed using site characteristics which resulted in five subjective homogenous regions. To check homogeneity, discordancy and heterogeneity measures were implied based on at-site statistics of the stations using L-moments approach. These statistical measures established validity of five HCR for Pakistan and showed that elevation has a key role in the construction of regions. Best-fit distributions were selected in the form of Generalized Pareto distribution and Pearson type-III distribution using L-Moment Ratio Diagram (LMRD) and goodness-of-fit (GOF) test for drought risk assessment. Projections of drought were obtained for each region in the form of regional quantiles at various return periods using regional frequency analysis (RFA). Finally, at-site frequency analysis (ASFA) was performed to calculate quantiles at different return periods and project drought at each individual site using best-fit distributions. The uncertainty of both types of quantile results was assessed using Monte Carlo Simulation which showed similarity at lower while increasing uncertainty at higher return periods. From our analysis, we found that RFA is statistically better due to many values from sites of the region as compared to ASFA for future drought risk assessment. According to the results, regions one and two are enriched of water resources due to rainfall and glaciers, region five has enough rain in monsoon season while regions three and four include most vulnerable and drought-prone areas of Sindh and Baluchistan provinces and have chances of severe to extreme drought in the near future. The findings of this study are helpful for policymakers, monitoring droughts and its projections, water resource planners, etc. in the study area.

A hybrid method for forecasting new product sales based on fuzzy clustering and deep learning

Purpose The purpose of this paper is to solve the problem of low accuracy in new product demand forecasting caused by the absence of historical data and inadequate consideration of influencing factors. Design/methodology/approach A hybrid new product demand forecasting model combining clustering analysis and deep learning is proposed. Based on the product similarity measurement, the weight of product similarity attributes is realized by using the method of fuzzy clustering-rough set, which provides a basis for the acquisition and collation of historical sales data of similar products and the determination of product similarity. Then the prediction error of Bass model is adjusted based on similarity through a long short-term memory neural network model, where the influencing factors such as product differentiation, seasonality and sales time on demand forecasting are embedded. An empirical example is given to verify the validity and feasibility of the model. Findings The results emphasize the importance of considering short-term impacts when forecasting new product demand. The authors show that useful information can be mined from similar products in demand forecasting, where the seasonality, product selling cycles and sales dependencies have significant impacts on the new product demand. In addition, they find that even in the peak season of demand, if the selling period has nearly passed the growth cycle, the Bass model may overestimate the product demand, which may mislead the operational decisions if it is ignored. Originality/value This study is valuable for showing that with the incorporation of the evaluation method on product similarity, the forecasting model proposed in this paper achieves a higher accuracy in forecasting new product sales.

Assets, Rural Livelihood Strategies and Welfare Outcomes: A Case Study from South Assam, India

This article uses income and asset data from a sample of 394 rural households from the remote and agriculturally backward region of the Barak Valley, located in the southern part of the north-eastern state of Assam in India, to construct a poverty profile for the area based on the asset structure of households. In doing so, the study goes beyond the traditional income-centric approach which treats the poor as a homogeneous group and takes a deeper look into the character of poverty in the region by segregating households on the basis of various poverty configurations. Further, using a combination of cluster analysis and propensity score matching methods, the study explores the relation between asset ownership and the choice of livelihood strategies and subsequently evaluates the impact of such choices on the income poverty status of rural households. The study finds that vulnerability is very high in the study area, with two-thirds of the rural households being afflicted with one type of deprivation or the other. However, the nature of deprivation being diverse, the study underscores the need for profiling poverty in terms of its structural and stochastic components and designing policy interventions suited to specific situations, for maximisation of welfare gains.

Molecular characterization of Sarocladium oryzae isolates causing sheath rot in paddy (Oryza sativa)

Sheath rot, caused by Sarocladium oryzae (Sawada) W. Gams & D. Hawksw, is an important seed borne fungal disease of paddy (Oryza sativa L.). In the present investigation, diseased specimens were collected in kharif 2016 and 2017 from more than 55 different geographical locations spread across 15 states of the country and variability in 30 isolates of Sarocladium oryzae was studied by molecular methods. The molecular characterization using two marker systems, i.e. Inter-Simple Sequence Repeat (ISSR) and Random Amplified Polymorphic DNA (RAPD) was carried out which revealed differences in isolates collected from different geographical locations. After preliminary screening, out of 22 ISSR and 24 RAPD primers, 8 ISSR and 7 RAPD primers were used against all the isolates. Percent polymorphic ranging between 53.1–100% in ISSR and 61.4–100% in RAPD was recorded with all the isolates. The average number of bands in ISSR markers was 63.3, and 71.8 in RAPD markers. All RAPD primers showed 100% polymorphism except OPS13 (61.4%) and OPD5 (74%). Among ISSR primers ISSR5 and ISSR18 showed 100% polymorphism. Cluster analysis of individual and combined primers was done. In the combined cluster analysis of RAPD and ISSR markers, the isolates from same geographical region were in same cluster as evident from the cluster I that includes Kadapa, Ragolu, Maruteru, West Godavari and Adutharai, which are from southern plain region of the country. Similar observations were recorded when data analysis of individual marker system was carried out which revealed that the two different marker systems applied complement each other in characterizing various isolates.

Representation of modern pollen assemblages with respect to vegetation and climate in Northeast China

The characterization of modern pollen–vegetation–climate relationships forms the basis of past environmental and climatic reconstructions from fossil pollen records. Northeast China, which is located in a high-latitude temperate zone, is an ideal area for investigating the relationship between modern pollen assemblages and climate. However, a scarcity of datasets and detailed research on surface pollen in Northeast China prohibits high-resolution quantitative climate reconstructions in China. Here, a total of 43 surface moss samples from forest regions in the eastern part of Northeast China were collected and analyzed for pollen content. The pollen-vegetation-climate relationship was investigated by calculating distance-weighted vegetation abundances, cluster analysis, indicator species analysis, ordination analysis, correlation analysis and biomization. Pollen assemblages were compared with vegetation types recorded at the sampling sites on different scales. The combination of cluster analysis, indicator species analysis and biomization identified which forest types the surface pollen assemblages represented. Our results showed that: 1) Surface pollen assemblages from the forest region of Northeast China can be used to distinguish mixed forest types from coniferous forest at a regional scale. Indicator taxa of coniferous forest include Alnus and Betula, while the indicator taxa of mixed forest are the combinations of Juglans, Tilia, Ulmus, Quercus and Pinus. 2) Principal Component Analysis (PCA) indicates that the pollen assemblages are mainly controlled by temperature and precipitation, while the mean temperature of the coldest month (Mtco) is the most significant factor controlling vegetation change. 3) The Betula/Quercus ratio drops dramatically from coniferous forest to mixed forest with an increasing Mtco and can thus be used as an indicator of vegetation types at a regional scale and as an index to reconstruct temperature.

Наукометрический анализ сельскохозяйственного научного направления

Abstract. There is a wide range of article and literature in agricultural sciences, but only a few review of existing studies has systematically analyzed and visualized the trends in agricultural research. This study reviews agricultural research mainly published in the Scopus and Web of Science databases from 2013 to 2018 with the scientometric analysis. The main hot Research Fronts mainly cover research on plant-gene regulatory networks and genome editing; crop disease and pest control; food nutrition and safety; photosynthesis; plant rhizosphere microbial community; immunity of aquatic animals; and forest tree cultivation. The analysis covers 325 966 scientific research in General Agricultural and Biological Sciences in online tool SciVal (Scopus). Most frequently occurred keywords indexed in the research area are patients, cells, humans. Three subject area were identified with the combination of cluster analysis with VosViewer: „Medicine“, „Biochemistry, Genetic and Molecular Biology“, „Engineering“. Nearly half the papers (49,3 %) were published by the following three countries: the United States, China and the United Kingdom. Russia's contribution to the global ranking is 1,8 %. The most popular journals by average number of publication produced in the United Kingdom. Web of Science data analyzed with a tool InCites. The analysis covers 141 010 papers in categories „Agricultural Economics & Policy“, „Agricultural Engineering“, „Agriculture, Dairy & Animal Science“, „Agriculture, Multidisciplinary“. Сompilation of bibliometric maps allows to identify the commonly used keywords: evidence, pretreatment, cattle, growth. Four research themes were identified with the combination of cluster analysis: „Material Science Multidisciplinary“, „Biochemistry and Molecular Biology“, „Engineering Electrical Electronic“, „Chemistry, Physical“. In the Web of Science, the most cited journals have been taken by the company „Elsiever“. The study revealed that the Scopus is a more developed database on the direction of agriculture. The potential of Russian journals is determined by sampling journals in the RSCI by the number of citations. This review provides an in-depth understanding of existing agricultural trends in this research domain.

Exploration during early life: distribution, habitat and orientation preferences in juvenile king penguins

BackgroundThe early life of marine apex predators is poorly known, particularly for diving species. The orientation and foraging skills are presumably less developed in juveniles than in adults, especially during their first year at sea when juveniles might disperse further than adults.MethodsOver two years of monitoring, we tracked the movements of 17 juvenile king penguins (Aptenodytes patagonicus, ~ 1 year old) using satellite relay tags from Crozet Archipelago (Southern Indian Ocean), starting when birds left their natal colony for the first time. For comparison we also tagged 6 non-breeding adults, which at that stage, similar to juveniles, are unhampered by reproductive constraints and might roam further than breeders. We used a combination of cluster analysis and habitat modelling to investigate and compare the movement patterns and habitat use of experienced (non-breeding adults) and non-experienced (juveniles) individuals.ResultsWhile juvenile penguins and non-breeding adults followed similar routes, the movements by adults started later in the season and ranged over a considerably smaller area than juveniles. Net squared displacement analysis revealed that both groups did not move to a specific wintering area. Changes in direction of juveniles in respect to their departure island were similar and synchronous for both years. Habitat models revealed that foraging behaviour was affected by environmental variables such as wind or current speeds, sea surface temperature, or oceanic productivity, for both stages. Analysis of tracks revealed that birds moved predominately perpendicular or against the main direction of the Antarctic Circumpolar Current and the prevailing wind during austral summer (juveniles only) and autumn (juveniles and non-breeding adults). However, both juveniles and adults were more likely to move against the prevailing winds if productivity increased along their trajectories.ConclusionsThe exceptional duration of our tracking study provided unprecedented insights into the distribution, habitat preferences and orientation of two poorly known life history stages of an expert avian diver. Our study suggests that juveniles might use both innate and learnt skills to reach profitable foraging areas during their first year at sea, which is critical in long-lived species.

Proposition of Machine Learning Driven Personalized Marketing Approach for E-Commerce

The ubiquity of digital devices and Internet has formed a constantly connected online environment which led to the extensive adoption of e-commerce. However, the active participation of growing number of stakeholders intensifies the highly competitive landscape of the dynamic e-commerce market and the scarcity of trust in e-commerce business impede the generation of consistent sales growth. The obstruction necessitates the implementation of innovative marketing strategies to enhance the relationships with customers to develop customer loyalty. Therefore, a machine learning driven personalized marketing approach is proposed to facilitate the implementation of personalized marketing in which there are 2 significant sequential elements namely, the development of personalized marketing contents and delivery of the contents to prospective customers. Cluster analysis is employed to perform customer segmentation to discover customer segments due to the capability of the analysis to identify similarities in customer preferences in which the discovered customer segments are used to construct personalized marketing contents. In addition, artificial neural network is employed to predict prospective customers due to the capability of artificial neural network to comprehend complex relationships between customer demographics and buying behaviour in which the prediction facilitates the delivery of the constructed personalized marketing contents to potential repeat customer to optimize the marketing initiative. The combination of cluster analysis and artificial neural network empowers the construction of an efficacious marketing pipeline which enhances the competency of e-commerce businesses.

Construction of homogeneous climatic regions by combining cluster analysis and L‐moment approach on the basis of Reconnaissance Drought Index for Pakistan

AbstractDrought is a natural hazard which has slow onset and assessment of its vulnerability is important for water resources management and sustainable development particularly in the most vulnerable areas to climate change and its worse impacts. Pakistan is facing acute water scarcity and regionalization of drought is a pre‐planning strategy for coping the severe condition of water management in the country. This study is conducted to identify homogenous climatic regions (HCRs) with respect to drought condition based on Reconnaissance Drought Index using climate data of 55 metrological stations across Pakistan. For this purpose two statistical methods, cluster analysis and L‐moments technique were used. First, cluster analysis was performed using site characteristics of the stations which identified five homogenous groups. Second, two statistical measures that is, discordancy measures and heterogeneity measure were implied after some regional adjustments, to satisfy the homogeneity of the homogenous groups developed in step one based on at‐site statistics of the stations using L‐moments technique. These statistical measures established validity of five homogeneous climatic regions for Pakistan. Third, to assess these HCRs several results using site characteristics and climate data of the stations were calculated which show that elevation has key role in the rainfall pattern and aridity of the regions. According to the results, there is high variability among these regions with respect to water recourses. First two regions consists most elevated areas with enough rainfall and glacier. Regions (3) and (4) include most vulnerable and drought‐prone areas of Sindh and Baluchistan provinces while Region (5) has enough rainfall in monsoon season. This study findings will help in adaptation, mitigation and future's planning regarding climate, water resources management, monitoring droughts and its projections etc. in different areas and fill the gaps in existing literature in Pakistan.

E-government recommendation algorithm based on probabilistic semantic cluster analysis in combination of improved collaborative filtering in big-data environment of government affairs

In order to solve the problem of information overload and satisfy their individual needs in the use of e-government system, a personalized e-government recommendation algorithm combining probabilis...

Combined XRD-XRF cluster analysis for automatic chemical and crystallographic surface mappings

X-ray diffraction-X-ray fluorescence (XRD-XRF) data sets obtained from surface scans of synthetic samples have been analysed by means of different data clustering algorithms, with the aim to propose a methodology for automatic crystallographic and chemical classification of surfaces. Three data clustering strategies have been evaluated, namely hierarchical, k-means, and density-based clustering; all of them have been applied to the distance matrix calculated from the single XRD and XRF data sets as well as the combined distance matrix. Classification performance is reported for each strategy both in numerical form as the corrected Rand index and as a visual reconstruction of the surface maps. Hierarchical and k-means clustering offered comparable results, depending on both sample complexity and data quality. When applied to XRF data collected on a two-phases test sample, both algorithms allowed to obtain Rand index values above 0.8, whereas XRD data collected on the same sample gave values around 0.5; application to the combined distance matrix improved the correlation to about 0.9. In the case of a more complex multi-phase sample, it has also been found that classification performance strongly depends on both data quality and signal contrast between different regions; again, the adoption of the combined dissimilarity matrix offered improved classification performance.

An integrated analysis of hyperspectral and morphological data of cicada ovipositors revealed unexplored links to specific oviposition hosts

Oviposition host selection by females greatly influences the survival of insect offspring. Based on data collected from 11 cicada species, we propose a new analytical approach to examine associations between oviposition host selection (live or dead wood) and three types of quantitative data (reflectance data acquired from ovipositors, morphometric data, and types and distribution of various sensilla). The study hypothesis was that a combination of these quantitative data would provide accurate characterization of species with known oviposition hosts and also enable prediction of a species with unknown oviposition hosts. Using cluster analysis based on morphometric and types and distribution of various sensilla as explanatory variables, the 11 cicada species were clearly categorized into two groups, i.e., one group ovipositing in live twigs, and the other group ovipositing in dead twigs. Although host selection of two species with unknown ovipositional preference awaits to be confirmed by future field investigation, we used the proposed integrated functional morphology approach to predict that these two species oviposit in dead twigs. Animal ecologists and evolutionary biologists frequently face the challenge of having incomplete information about function and behavior of animals, and therefore, seek to perform functional morphological studies. The combination of cluster analysis of morphometric and reflectance data as explanatory variables described in this study may be considered of considerable relevance to this broad audience.

A scientometric analysis and critical review of construction related ontology research

Abstract There is a wide range of literature on adopting ontology to solve construction problems, but no review of existing studies has systematically analyzed and visualized the trends in ontology research. This study reviews ontology research mainly published in the Scopus database from 2007 to 2017 with the combination of scientometric analysis and critical review. Scientometric analysis (e.g. co-author, co-word, co-citation, and clusters) objectively visualized the research status quo while a critical review was used to identify the research themes and challenges of ontology research in the construction industry. The results identified a large network of co-authors in this field to understand collaboration relationships. Over half the papers (53%) were published by the following three countries: the United States, the United Kingdom, and Canada. The top co-occurring keywords were “project management” at which ontology facilitates knowledge management and information retrieval. When the time factor was taken into consideration, keywords naturally evolved from “project management”, and “knowledge management” to “building information modeling”, and “compliance control” with the successful adoption of information techniques in the construction industry. Four research themes were identified with the combination of cluster analysis and critical review: “Domain ontology”, “Industry foundation classes”, “Automated compliance checking”, and “Building information modeling”. This review provides an in-depth understanding of existing ontology research and indicates the emerging trends in this research domain.

Lithium-Ion Battery Packs Formation With Improved Electrochemical Performance for Electric Vehicles: Experimental and Clustering Analysis

With the increase of production of electrical vehicles (EVs) and battery packs, lithium ion batteries inconsistency problem has drawn much attention. Lithium ion battery imbalance phenomenon exists during three different stages of life cycle. First stage is premanufacturing of battery pack i.e., during the design, the cells of similar performance need to be clustered to improve the performance of pack. Second is during the use of battery pack in EVs, batteries equalization is necessary. In the third stage, clustering of spent lithium ion batteries for reuse is also an important problem because of the great recycling challenge of lithium batteries. In this work, several clustering and equalization methods are compared and summarized for different stages. The methods are divided into the traditional methods and intelligent methods. The work also proposes experimental combined clustering analysis for new lithium-ion battery packs formation with improved electrochemical performance for electric vehicles. Experiments were conducted by dismantling of pack and measurement of capacity, voltage, and internal resistance data. Clustering analysis based on self-organizing map (SOM) neural networks is then applied on the measured data to form clusters of battery packs. The validation results conclude that the battery packs formed from the clustering analysis have higher electrochemical performance than randomly selected ones. In addition, a comprehensive discussion was carried out.

Adaptation of invasive species of Solidago canadensis L. in the urban ecosystem of Cherkasy

Introduction . There is a great deal of researches on the effects of S. canadensis L. on species of phytocenoses and biodiversity in general, however the results of such researches have many contradictions. They differ depending on their research locations . This proves that the behavior of S. canadensis L. in phytocenoses is highly dependent on many factors (soil, climate, species composition of local phytocenoses ) as well as the lack of a clear understanding of the mechanisms of its impact on phytocenoses, indicating the need to further study the impact of this species on phytocenoses to understand this mechanism. Purpose. The detection of S. canadensis L . localities in the Cherkasy urban ecosystem, study of synecological parameters of S. canadensis L. and accompanying species, as well as its influence on the floristic composition of phytocenoses; predicting the possibility of further increasing the invasion area and identifying priority areas for measures to control it. Methods. Locations were searched using the route method. The accounting was carried out at accounting sites of 1 m 2 , using standard geobotanical techniques. Statistical processing of the results was performed using dispersion coefficient, diversity compositioт coefficient, occurrence coefficient, Jaccard similarity coefficient and cluster analysis by nearest neighbor method. Results. In the course of the research conducted during the vegetation period 2016-2017, a significant level of presence of the S. canadensis L . was noted in 9 districts in Cherkasy: average projective coverage of 20,00% - 82 , 67%. S. canadensis L . dominates on 74% of the accounting sites. In all districts depletion of the species composition of spontaneous plant communities due to the invasion of the S. canadensis L. was found: the average number of species per 1 m 2 was 6.68 ± 0.23 species. The occurrence coefficients of the accompanying species are calculated, the species occurring in phytocenoses with S. canadensis L . are most often identified: Elymus repens (L.) Gould (76%) Ambrosia artemisiifolia L . (64%) Achillea submillefolium Klok. et Krytzka (50%) Stenactis annua (L.) Cass. ex Less. (36%) Acer negundo L. (30%) Sonchus arvensis L. (26%) Poa trivialis L . (20%) Trifolium pratense L . (18%) Artemisia absinthium L. (16%) Trifolium arvense L . (16%), Artemisia vulgaris L. (14%). In comparing the species composition of the phytocenoses of the 9 districts by Jaccard similarity coefficient and the cluster analysis of the average projective coverage of the S. canadensis L. and accompanying species, priority areas were identified to control the invasion of the S. canadensis L . Combining cluster analysis with Jaccard similarity coefficient increases the accuracy of the results (because the species composition of the phytocenoses in addition to their quantitative indicators is taken into account). Similar invasion behavior is predicted in areas with similar species composition of phytocenoses. Conclusion. S. canadensis L . was detected in 9 Cherkasy districts. Negative influence of the S. canadensis L . on the species composition of phytocenoses has been proved. The species that were found most often with the S. canadensis L . have adapted to the conditions it creates. The combination of Jaccard similarity coefficient and cluster analysis yields more accurate results than cluster analysis alone.

Efficiency and effectivity of high precision grinding manufacturing processes: An approach based on combined DEA and cluster analyses

The manufacturing reproducibility of high precision grinding patterns depends on many process influences and technology characteristics, like cutting speed, lining, grinding disc, cutting fluid and longitudinal feed. This paper outlines a combined approach for the detection of grinding parameter settings, which ensure on the one hand reproducibility and on the other hand guarantee economy. Cluster analytics is utilised to identify grinding parameter sets and grinding results with similar characteristics within a group. Data Envelopment Analysis (DEA) is applied to describe the efficiency of these grinding processes based on input and output parameters which are selected with regard to grinding patterns.

Under-sampling class imbalanced datasets by combining clustering analysis and instance selection

Class-imbalanced datasets, i.e., those with the number of data samples in one class being much larger than that in another class, occur in many real-world problems. Using these datasets, it is very difficult to construct effective classifiers based on the current classification algorithms, especially for distinguishing small or minority classes from the majority class. To solve the class imbalance problem, the under/oversampling techniques have been widely used to reduce and enlarge the numbers of data samples in the majority and minority classes, respectively. Moreover, the combinations of certain sampling approaches with ensemble classifiers have shown reasonably good performance. In this paper, a novel undersampling approach called cluster-based instance selection (CBIS) that combines clustering analysis and instance selection is introduced. The clustering analysis component groups similar data samples of the majority class dataset into ‘subclasses’, while the instance selection component filters out unrepresentative data samples from each of the ‘subclasses’. The experimental results based on the KEEL dataset repository show that the CBIS approach can make bagging and boosting-based MLP ensemble classifiers perform significantly better than six state-of-the-art approaches, regardless of what kinds of clustering (affinity propagation and k-means) and instance selection (IB3, DROP3 and GA) algorithms are used.

A machine learning approach relating 3D body scans to body composition in humans.

A long-standing question in nutrition and obesity research involves quantifying the relationship between body fat and anthropometry. To date, the mathematical formulation of these relationships has relied on pairing easily obtained anthropometric measurements such as the body mass index (BMI), waist circumference, or hip circumference to body fat. Recent advances in 3D body shape imaging technology provides a new opportunity for quickly and accurately obtaining hundreds of anthropometric measurements within seconds, however, there does not yet exist a large diverse database that pairs these measurements to body fat. Herein, we leverage 3D scanned anthropometry obtained from a population of United States Army basic training recruits to derive four subpopulations of homogenous body shape archetypes using a combined principal components and cluster analysis. While the Army database was large and diverse, it did not have body composition measurements. Therefore, these body shape archetypes were paired to an alternate smaller sample of participants from the Pennington Biomedical Research Center in Baton Rouge, LA that were not only similarly imaged by the same 3D scanning machine, but also had concomitant measures of body composition by dual-energy X-ray absorptiometry body composition. With this enhanced ability to obtain anthropometry through 3D scanning quickly of large populations, our machine learning approach for pairing body shapes from large datasets to smaller datasets that also contain state-of-the-art body composition measurements can be extended to pair other health outcomes to 3D body shape anthropometry.

Examining driver injury severity in intersection-related crashes using cluster analysis and hierarchical Bayesian models

Traffic crashes are more likely to occur at intersections where the traffic environment is complicated. In this study, a hybrid approach combining cluster analysis and hierarchical Bayesian models is developed to examine driver injury severity patterns in intersection-related crashes based on two-year crash data in New Mexico. Three clusters are defined by K-means cluster analysis based on weather and roadway environmental conditions in order to reveal drivers’ risk compensation instability under diverse external environment. Hierarchical Bayesian random intercept models are developed for each of the three clusters as well as the whole dataset to identify the contributing factors on multilevel driver injury outcomes: property damage only (Level I), complaint of injury and visible injury (Level II), and incapacitating injury and fatality (Level III). Model comparison with an ordinary multinomial logistic model omitting crash data hierarchical features and cross-level interactions verifies the suitability and effectiveness of the proposed hybrid approach. Results show that a number of crash-level variables (time period, weather, light condition, area, and road grade), vehicle/driver-level variables (traffic controls, vehicle action, vehicle type, seatbelt used, driver age, drug/alcohol impaired, and driver age) along with some cross-level interactions (i.e., left turn and night, drug and dark) impose significantly influence driver injury severity. This study provides insightful understandings of the effects of these variables on driver injury severity in intersection-related crashes and beneficial references for developing effective countermeasures for severe crash prevention.