Two-step Clustering Method Research Articles

A two-step clustering method for facility location problem

Facility location problems are designed with the objective to gain more profit. The profit can be gained when the maximum demand is satisfied. The demand can be satisfied when maximum number of customers are covered or served. To attain maximum number of customers, various approaches have been investigated. In general, most of the approaches consider for the facility location models based on radius as a service area of facility. Therefore, such facilities which fulfil their service in a radius can be served by conventional approach. However, conventional approaches fail to allocate those facilities which are not inclined by topographical and road network barriers. In this paper, we propose a model to optimise facility allocation in such scenarios. In the propose model, we have used a two-step clustering approach to solve the facility location problem. Experimental results illustrate that the proposed algorithm based on density affinity propagation (DAP) for the facility location problem can be used to construct a solution for maximal service and covering area.

Customer relationship management analysis of outpatients in a Chinese infectious disease hospital using drug-proportion recency-frequency-monetary model

BackgroundIdentifying the patient types with different economic values can be useful for hospital development. ObjectiveThis work uses the theory of customer relationship management (CRM) to analyze the outpatients in the hospital for infectious diseases in Shanghai, China. MethodsA total of 2,271,020 data elements of outpatients in the research unit between August 2009 and December 2019 were extracted, analyzed and cleaned to obtain 171,107 valid data elements (1 element per person). The main diseases were viral hepatitis B (VHB) and acquired immunodeficiency syndrome (AIDS), and the average percentage of drug expenditure was 80.39 %.We innovatively expanded the classic RFM (R: recency, F: frequency, M: monetary) model in CRM to the dRFM (d: percentage of drug expenditure) model. We selected the best clustering algorithm from the K-means, Kohonen and two-step clustering methods to find the optimal model to distinguish the types of patients with different economic values and the best decision-making algorithm from the C5.0, CART classification regression tree, CHAID and QUEST algorithms to verify the model. ResultsAfter performing two rounds of K-means clustering analysis on three models: RFM, RFM + dRFM and dRFM, and 97,855 data elements were retained. The RFM + dRFM model was the optimal model, clustering the patients into 3 types: potential patients (24.2 %) to be retained, with a high drug expenditure and the last visit in more than 19.06 months, high-value patients (24.5 %) to be attracted, with the last visit in about 6.66 months; basal patients (51.3 %) to be kept, with the last visit in about 3.7 months. The model was then verified using the C5.0 decision tree algorithm with an accuracy rate of 99.97 %. ConclusionThis objective CRM analysis of the patients in the hospital for infectious diseases using the dRFM model accurately identified different types of patients, providing an objective and effective basis for hospital management.

Diversity and Resilience to Socio-Ecological Changes of Smallholder Lagune Cattle Farming Systems of Benin

The indigenous smallholder Lagune breed and the production systems in which it is embedded in Southern Benin have received very little research or policy attention. Consequently, very little information exists on the diversity of these production systems or on their capacity to adapt to ongoing socio-economic and environmental changes. This study aimed to explore and characterize the diversity of Lagune cattle production systems along with farmers’ local knowledge and resilience strategies. A questionnaire was administered to 417 Lagune cattle farmers across two agro-ecological zones: Ouémé Valley (OVZ) and Pobe (PZ). It included, inter alia, questions related to households’ socio-economic conditions, their cattle herd characteristics, and management practices. Categorical principal component analysis and the two-step clustering method were used to classify the production systems which were then compared using the chi-square and ANOVA procedures. Four distinct farm types were identified. This study revealed the important role of agroecology in the diversity of farmers’ breeding practices. Controlled mating was more common in tethering systems whereas uncontrolled mating, widespread in free-roaming systems, has favored Lagune breed admixture with zebus. Opportunities for conserving the genetic diversity within the Lagune breed might be greater in PZ where breed admixture was almost inexistent.

Unsupervised Event Detection, Clustering, and Use Case Exposition in Micro-PMU Measurements

Distribution-level phasor measurement units, a.k.a, micro-PMUs, report a large volume of high resolution phasor measurements which constitute a variety of event signatures of different phenomena that occur all across power distribution feeders. In order to implement an event-based analysis that has useful applications for the utility operator, one needs to extract these events from a large volume of micro-PMU data. However, due to the infrequent, unscheduled, and unknown nature of the events, it is often a challenge to even figure out what kind of events are out there to capture and scrutinize. In this paper, we seek to address this open problem by developing an unsupervised approach, which requires minimal prior human knowledge. First, we develop an unsupervised event detection method based on the concept of Generative Adversarial Networks (GAN). It works by training deep neural networks that learn the characteristics of the normal trends in micro-PMU measurements; and accordingly detect an event when there is any abnormality. We also propose a two-step unsupervised clustering method, based on a novel linear mixed integer programming formulation. It helps us categorize events based on their origin in the first step and their similarity in the second step. The active nature of the proposed clustering method makes it capable of identifying new clusters of events on an ongoing basis. The proposed unsupervised event detection and clustering methods are applied to real-world micro-PMU data. Results show that they can outperform the prevalent methods in the literature. These methods also facilitate our further analysis to identify important clusters of events that lead to unmasking several use cases that could be of value to the utility operator.

Data mining application to healthcare fraud detection: a two-step unsupervised clustering method for outlier detection with administrative databases.

BackgroundThe healthcare sector is an interesting target for fraudsters. The availability of a great amount of data makes it possible to tackle this issue with the adoption of data mining techniques, making the auditing process more efficient and effective. This research has the objective of developing a novel data mining model devoted to fraud detection among hospitals using Hospital Discharge Charts (HDC) in Administrative Databases. In particular, it is focused on the DRG upcoding practice, i.e., the tendency of registering codes for provided services and inpatients health status so to make the hospitalization fall within a more remunerative DRG class.MethodsWe propose a two-step algorithm: the first step entails kmeans clustering of providers to identify locally consistent and locally similar groups of hospitals, according to their characteristics and behavior treating a specific disease, in order to spot outliers within this groups of peers. An initial grid search for the best number of features to be selected (through Principal Feature Analysis) and the best number of local groups makes the algorithm extremely flexible. In the second step, we propose a human-decision support system that helps auditors cross-validating the identified outliers, analyzing them w.r.t. fraud-related variables, and the complexity of patients’ casemix they treated. The proposed algorithm was tested on a database relative to HDC collected by Regione Lombardia (Italy) in a time period of three years (2013-2015), focusing on the treatment of Heart Failure.ResultsThe model identified 6 clusters of hospitals and 10 outliers among the 183 units. Out of those providers, we report the in depth the application of Step Two on three Hospitals (two private and one public). Cross-validating with the patients’ population and the hospitals’ characteristics, the public hospital seemed justified in its outlierness, while the two private providers were deemed interesting for a further investigation by auditors.ConclusionsThe proposed model is promising in identifying anomalous DRG coding behavior and it is easily transferrable to all diseases and contexts of interest. Our proposal contributes to the limited literature regarding behavioral models for fraud detection, identifying the most ’cautious’ fraudsters. The results of the first and the second Steps together represent a valuable set of information for auditors in their preliminary investigation.

PENGIDENTIFIKASIAN SEGMENTASI PENGGUNA SISTEM MANAJEMEN PEMBELAJARAN SEBUAH UNIVERSITAS DENGAN METODE TWO-STEP CLUSTERING

This study aims to identify user segmentations of a learning management system at Universitas Negeri Malang using two-step clustering method. Data is collected through an on-line survey which is linked onto the academic information systems. Total number of data are 10.594 responses. This study runs three clustering methods, two-step, k-means, and kohonen clustering, in order to identify the user segmentation. Two-step clustering methods performs better than the other two clustering methods (k-means and kohonen) based on silhouette clustering index. The number of cluster resulted from two-step clustering method are three clusters. The top three of most important factors in identifying the user segmentation are the benefit of the learning management usage in helping to access information, increasing the effectiveness of accomplishing assignments, and students satisfaction on the learning management system services.

An OD Flow Clustering Method Based on Vector Constraints: A Case Study for Beijing Taxi Origin-Destination Data

Origin-destination (OD) flow pattern mining is an important research method of urban dynamics, in which OD flow clustering analysis discovers the activity patterns of urban residents and mine the coupling relationship of urban subspace and dynamic causes. The existing flow clustering methods are limited by the spatial constraints of OD points, rely on the spatial similarity of geographical points, and lack in-depth analysis of high-dimensional flow characteristics, and therefore it is difficult to find irregular flow clusters. In this paper, we propose an OD flow clustering method based on vector constraints (ODFCVC), which defines OD flow event point and OD flow vector to express the spatial location relationship and geometric flow behavior characteristics of OD flow. First, the OD flow vector coordinate system is normalized by the Euclidean distance-based OD flow event point spatial clustering, and then the OD flow clusters with similar flow patterns are mined using adjusted cosine similarity-based OD flow vector feature clustering. The transformation of OD data from point set space to vector space is realized by constraining the vector coordinate system and vector similarity through two-step clustering, which simplifies the calculation of high-dimensional similarity of OD flow and helps mining representative OD flow clusters in flow space. Due to the OD flow cluster property, the k-means algorithm is selected as the basic clustering logic in the two-step clustering method, and a sum of squared error perceptually important points algorithm considering silhouette coefficients (SSEPIP) is adopted to automatically extract the optimal cluster number without defining any parameters. Tested by origin-destination flow data in Beijing, China, new traffic flow communities based on traffic hubs are obtained by using the ODFCVC method, and irregular traffic flow clusters (including cluster mode, divergence mode, and convergence mode) with representative travel trends are found.

Profile of service providers in high-risk neonatal care in the databases of the Brazilian Unified National Health System

The study aimed to analyze the profile of service providers in the Brazilian Unified National Health System (SUS) in relation to high-risk neonatal care procedures, based on data from the public-access database on Authorizations for Hospital Admissions (AIH) without identification of the neonatal patient. This cross-sectional study used data on AIH of unidentified newborns up to one day old, issued from January 2013 to December 2015, and from the National Registry of Healthcare Establishments. The TwoStep Cluster method was used to classify the establishments with similar characteristics. Regional differences were observed in the frequency of use of neonatal intensive care units and in special procedures recorded in the databases. The TwoStep Cluster method identified three groupings of providers: the first with 1,151 establishments and with low-complexity hospitalizations, the second with 84 establishments and surgical-profile hospitalizations, and the third with 393 establishments and higher complexity hospitalizations in neonatal care, including surgeries. The records on admissions jointly with multivariate analytical techniques have the potential to support decision-making by SUS administrators in the organization of neonatal care.

Classifications of Stations in Urban Rail Transit based on the Two-step Cluster

Different types of stations have different functional roles in the urban rail transit network. Firstly, based on the characteristics of the urban rail transit network structure, the time series features and passenger flow features of the station smart card data are extracted. Secondly, we use the principal component analysis method to select the suitable clustering variables. Finally, we propose a station classification model based on the two-step cluster method. The effectiveness of the proposed method is verified in the Beijing subway. The results show that the proposed model can successfully identify the types of urban rail transit stations, clarify the function and orientation of each station.

A Novel Technique for Eliminating Parameter Initialization in Clustering

This study revealed the results of simulating the techniques for eliminate cluster initialization in clustering through series of test of different clustering techniques. Clusters of similar database schema from heterogeneous data sources was successfully clustered at an improved accuracy of 93.33%. The accuracy of clusters using self-organizing map neural network (SOM) was improved by implementing the two-step clustering method prior to SOM with an accuracy of 93.33 percent, an increase of 6.33% from the clustering result of SOM. Agglomerative clustering with hierarchical clustering are best combined to come-up with an automatic clustering task, thus, initialization was eliminated.
 Keywords: clustering, agglomerative clustering, hierarchical clustering, self-organizing map neural networks

Clustering for multi-dimensional data set: a case study on educational data

This paper present two-step cluster method to handle data sets in educational data mining. It can handle multi-dimentional metric data points especially in complex data sets. For a case study, we use tracer study data in order to assign a clustering of alumni based on the profile. The result of clustering will help the school to evaluate and improve the quality of its graduates.

Modeling of Wind Speeds Inside a Wind Farm With Application to Wind Farm Aggregate Modeling Considering LVRT Characteristic

The wind speeds of each wind turbine generator (WTG) are the basis for establishing an aggregated model of a wind farm for power system transient simulation. However, the commonly obtained operation data of the wind farm do not include the wind speeds of each WTG. Thus, a modeling method for the wind speeds of the WTGs of same moment inside the wind farm was proposed. A wind speed combination model (WSCM) is established via K-means clustering from the sorted field-measured wind speed data, which can provide reasonable wind speeds for each WTG when aggregate modeling the wind farm. Subsequently, an improved two-step clustering method of the WTGs was proposed, by which the WTGs are initially clustered according to whether they enter low voltage ride through (LVRT) mode or not. An “N-1 VS One” aggregated model and the binary search algorithm were used to quickly and correctly predict whether a WTG in the wind farm enters LVRT mode under a grid fault. The case study demonstrates the satisfactory performance of the WSCM and the necessity of clustering the WTGs according to whether they enter or do not enter LVRT mode for obtaining an accurate aggregated model of the wind farm. All the data, models, and programs that are used in this paper can be found at IEEE Dataport, DOI: 10.21227/dqyk-t852.

Do residential location effects on travel behavior differ between the elderly and younger adults?

The built environment affects individuals’ travel behavior in a variety of dimensions, such as trip generation, mode choice, and travel duration. However, it is not well understood how these effects differ across different socioeconomic groups (e.g. the elderly versus younger adults) and how residential self-selection contributes to these differences. Using the 2013 Nanjing (China) Travel Survey data, this study estimates the differential responsiveness to the variation in residential location for different age groups. The two-step clustering method is applied to characterize two types of residential locations and the propensity score matching approach is utilized to address self-selection effects. We find that, after control for self-selection, residential location effects on travel behavior differ significantly between the elderly (60+ years old) and younger respondents (18–59 years old). Changes in the living environment play a more important role in influencing the elderly’s travel frequency and travel duration than those of younger adults. When we compare the observed effects of residential location, self-selection effects are modest for the elderly while they matter to a great extent for younger adults. In addition, due to differences in residential self-selection, there is an underestimation of residential location effects on the elderly’s travel behavior versus an overestimation of those for younger adults. These findings indicate that overlooking the variation of built environment effects between different age groups may lead to ineffective housing and transportation policy implications.

Graph clustering and variational image segmentation for automated firearm detection in X‐ray images

Millions of parcels and pieces of luggage are scanned daily for threat detection at border control or high-security buildings. Currently, the process is manually operated by security agents and it is slow and time-consuming. The automation of this process will lift the burden from the security agents and will allow larger volumes of items to be scanned. The authors consider the problem of automatic threat detection, in particular firearms, in X-rays. To achieve this goal, they propose a hybrid algorithm that combines two well-established image segmentation algorithms into a two-step clustering method. The first step is a semi-supervised spectral clustering algorithm at the image level, which classifies whole images into benign or containing a threat. The images classified as threatening from the first step proceed to the second stage, where a variational image segmentation algorithm performs clustering at the pixel level to locate the threat if it exists. The hybrid algorithm is designed to scale-up the processing of hundreds of images, in comparison to the academic literature where only a handful images are used for demonstration. Numerical experiments establish that the combination of two different algorithms produces better results than using individual algorithms.

Determinants of livelihood choice and implications for targeted poverty reduction policies: A case study in the YNL river region, Tibetan Plateau

Governmental policies that can develop advantageous endogenous livelihood transition are of great importance for anti-poverty. However, the existing research finds that in the Tibetan Plateau, there is a certain contradiction between the poverty alleviation policies and the local farmers’ needs, and thus the effect of these policies will be weakened and the goal of poverty alleviation be missed. It is therefore imperative to identify and analyze the factors influencing the livelihood choice so that policy makers and development practitioners can design policies that will tackle the poverty afflicting the majority of the local people. This paper provides a new method for anti-poverty policymaking and conducts a case study to enrich the safety net/cargo net paradigms. In this paper, household data in the YNL river region are used and a two-step clustering method is developed to classify these households into three types: hanging in (type I), stepping up (type II) and stepping out (type III). Then a multinomial logit model is adopted to identify determinants of households’ livelihood choice. The results indicate that drought poses a significant negative impact on farmers’ choice of type II, while household size, educational level of labor and medical expenditure have a significant positive impact on the choice of type II; spring frost, proportion of skilled training, borrowed money and the ratio of agricultural equipment have a significant positive impact on their choice of type III. Some anti-poverty policy implications are put forward in the last on the basis of these determinants.

A New Method of Privacy Protection: Random k-Anonymous

A new k-anonymous method which is different from traditional k-anonymous was proposed to solve the problem of privacy protection. Specifically, numerical data achieves k-anonymous by adding noises, and categorical data achieves k-anonymous by using randomization. Using the above two methods, the drawback that at least k elements must have the same quasi identifier in the k-anonymous data set has been solved. Since the process of finding anonymous equivalence is very time consuming, a two-step clustering method is used to divide the original data set into equivalence classes. First, the original data set is divided into several different sub-datasets, and then the equivalence classes are formed in the sub-datasets, thus greatly reducing the computational cost of finding anonymous equivalence classes. The experiments are conducted on three different data sets, and the results show that the proposed method is more efficient and the information loss of anonymous dataset is much smaller.

Firearms, Violence-Related Injuries, and Victimization Profiles: An Approach Using Cluster Analysis.

The aim of this study was to characterize the profile of nonlethal victims of urban violence by firearms and to describe traumas suffered by victims, according to a medical-legal and forensic perspective. A cross-sectional and exploratory study was conducted at the Center of Forensic Medicine and Dentistry in northeastern Brazil. The sample consisted of 233 victims of urban violence by firearm who presented some type of trauma. Descriptive and multivariate statistics using cluster analysis (CA) were performed. The TwoStep Cluster method was chosen to characterize the profile of victims. The night shift (56.8%) and the period corresponding to Saturdays (20.0%) and Sundays (20.4%) concentrated the largest number of occurrences. Cases of trauma in more than one region of the body simultaneously prevailed (31.8%). Based on the CA results, the formation of two clusters with distinct victimization profiles was verified. Cluster 1 was mostly characterized by younger single victims who suffered violence by firearm in the urban area perpetrated by an unknown perpetrator, resulting in greater occurrence of isolated upper and lower limb traumas. In contrast, Cluster 2 consisted essentially of older, married, or stable-union victims who experienced firearm violence in the suburban area, perpetrated by a known aggressor, resulting in greater occurrence of multiple traumas, that is, affecting several regions of the body at the same time. These findings reveal different risk groups for urban violence by firearms and traumas, contributing to the planning of strategies with emphasis on health care, prevention, and promotion.

Who are the big spending tourists travelling to South Africa?

Orientation: Tourism growth is not just about numbers anymore; it is about utilising the economic potential of every tourist visiting South Africa. This country needs to target the most lucrative markets to benefit economic growth. Research purpose: This article aims to identify the big spenders with reference to demographic characteristics and tourist destination preferences. Motivation for the study: The number of tourists to South Africa increases annually. However this is not evident in an increase in job opportunities or tourism products. The tourism industry also indicated that they receive less tourists. Therefore the increase in tourists do not necessarily lead to an increase in income which might relate to the type of tourist received. Research design, approach and method: A quantitative survey was done amongst international visitors leaving South Africa at O.R. Tambo International Airport, to which a two-step clustering method was applied. Main findings: Two spending groups were identified namely big spenders and average spenders. Big spenders are characterised by a specific demographic profile of being single, better qualified, in a professional occupation, and male. They tend to prefer visiting the iconic natural attractions where they spend significantly more than the average spenders. Practical/managerial implications: The implication of this study is that tourism bodies should focus more direct marketing efforts on the big spenders and provide products that suit the needs of this target market. Contribution/value-add: Effective market segmentation based on spending provide insight to markets that will create a high return on investment and this will directly contribute to the economic growth of the industry and alleviate poverty.

Cardiovascular disease lifestyle risk factors in people with psychosis: a cross-sectional study

BackgroundPeople with psychosis die on average 25 years earlier than those in the general population, with cardiovascular disease (CVD) contributing to much of the excess mortality. This cross-sectional study aimed to identify the relationship between lifestyle risk factors for CVD – poor nutrition, smoking and low physical activity levels – and dyslipidaemia, hypertension and hyperglycaemia while controlling for potential confounders in 1825 people from the Survey of High Impact Psychosis (SHIP) in Australia. We also aimed to identify clustering patterns of lifestyle risk factors and associated demographic variables.MethodsThree logistic regressions were used to predict the effect of nutrition, smoking and physical activity on dyslipidaemia, hypertension and hyperglycaemia while controlling for clozapine use, sex and age. Clustering patterns of nutrition, smoking and physical activity were examined using the two-step cluster method which is based on hierarchical cluster analysis. Demographic variables associated with different clusters were identified using measures of association.ResultsSmoking status had a positive association with dyslipidaemia (adjusted odds ratio = 0.50; 95% confidence interval = 0.32–0.78; p = 0.002). Other cardiovascular disease lifestyle risk factors did not have a significant relationship with dyslipidaemia, hypertension and hyperglycaemia. Clustering patterns of lifestyle risk factors showed that younger men, with low education levels, and relying on a government pension, were most likely to display the poorest lifestyle risk behaviours. The largest cluster (42%) of participants was characterised by a mixed demographic profile and were most likely to display poor nutrition and low physical activity levels but less likely to smoke.ConclusionsOnly smoking status had a significant positive association with dyslipidaemia which could indicate that there are additional factors affecting the relationship between other cardiovascular lifestyle risk factors and dyslipidaemia, hypertension and hyperglycaemia in people with psychosis. Unknown confounders and traditional lifestyle risk factors may explain the high rates of CVD in this group. Clustering of lifestyle risk factors and their demographic profiles could help the design of intervention programs in people with psychosis.

Exploring the labour productivity of agricultural systems across European regions: A multilevel approach

Agricultural productivity varies greatly among agricultural systems and among regions in Europe. A multilevel logistic regression model was applied to investigate the labour productivity of the six main agricultural systems across European NUTS2 regions. K-means and two-step clustering methods were used to classify European regions based on the agricultural systems’ standard output per annual work unit. We analysed the effect of environmental (soil erosion, rainfed yield potential), structural (farm education, age, pluriactivity, diversification, rented agricultural land), technical (yield ratio) and contextual (gross domestic product per capita, population density) factors on labor productivity. Significant differences were revealed between northern-central regions and the continental peripheries (Mediterranean, Eastern). Soil erosion negatively affects agricultural labour productivity; for each one ton/ha increase in the modelled annual soil erosion rate the odds of regions to attain high labour productivity decreased by 28%. The importance of technical efficiency in crop production was also identified. Observed low wheat yields, relative to modelled potential yields, in the southern, eastern and northern European regions indicate a large unexploited gap. The positive effect of the regional gross domestic product per capita and the low population density confirmed the importance of contextual factors on labour productivity. A second analysis of a composite indicator of the labour productivity of the European agricultural systems, which accounted for the productivity of each system without considering its size, revealed the positive effect of farm education and the negative effect of pluriactivity on agricultural labour productivity. The analyses indicate the importance of compensatory allowances for areas facing natural constraints, while investing in farm training schemes and advisory services could increase the adoption of new technologies and improve the performance of farmers in both economic and environmental terms. Finally, the significance of contextual factors indicates the importance of a better harmonisation of rural development policy with regional policy.