Search

The aim of the work is to identify a clustering structure for the 20 Italian regions according to the main variables related to COVID-19 pandemic. Data are observed over time, spanning from the last week of February 2020 to the first week of February 2021. Dealing with geographical units observed at several time occasions, the proposed fuzzy clustering model embedded both space and time information. Properly, an Exponential distance-based Fuzzy Partitioning Around Medoids algorithm with spatial penalty term has been proposed to classify the spline representation of the time trajectories. The results show that the heterogeneity among regions along with the spatial contiguity is essential to understand the spread of the pandemic and to design effective policies to mitigate the effects.

[Key message] The strong genetic structure observed in Mediterranean oats affects the predictive ability of genomic prediction as well as the performance of training set optimization methods., [Abstract] In this study, we investigated the efficiency of genomic prediction and training set optimization in a highly structured population of cultivars and landraces of cultivated oat (Avena sativa) from the Mediterranean basin, including white (subsp. sativa) and red (subsp. byzantina) oats, genotyped using genotype-by-sequencing markers and evaluated for agronomic traits in Southern Spain. For most traits, the predictive abilities were moderate to high with little differences between models, except for biomass for which Bayes-B showed a substantial gain compared to other models. The consistency between the structure of the training population and the population to be predicted was key to the predictive ability of genomic predictions. The predictive ability of inter-subspecies predictions was indeed much lower than that of intra-subspecies predictions for all traits. Regarding training set optimization, the linear mixed model optimization criteria (prediction error variance (PEVmean) and coefficient of determination (CDmean)) performed better than the heuristic approach “partitioning around medoids,” even under high population structure. The superiority of CDmean and PEVmean could be explained by their ability to adapt the representation of each genetic group according to those represented in the population to be predicted. These results represent an important step towards the implementation of genomic prediction in oat breeding programs and address important issues faced by the genomic prediction community regarding population structure and training set optimization., This work was supported by the Spanish Ministry of Science and Innovation [PID2019-104518RB-100], (AEI/FEDER, UE) and the regional government through the AGR-253 group and the European Regional and Social Development Funds. LGS is holder of a FPI fellowship from the Spanish Ministry of Economy and Competitiveness [BES-2017-080152]. This project has received funding from the European Union’s Horizon 2020 research and innovation program under grant agreement No 818144, and also the Severo Ochoa Program for Centres of Excellence in R&D. JIS was supported by the Beatriz Galindo Program (BEAGAL18/00115) from the Ministerio de Educación y Formación Profesional of Spain and the Severo Ochoa Program for Centres of Excellence in R&D from the Agencia Estatal de Investigación of Spain, grant SEV-2016-0672 (2017-2021) to the CBGP.

Background: Heterogeneity in Acute Respiratory Distress Syndrome (ARDS), as a consequence of its non-specific definition, has led to a multitude of negative randomised controlled trials (RCTs). Investigators have sought to identify heterogeneity of treatment effect (HTE) in RCTs using clustering algorithms. We evaluated the proficiency of several commonly-used machine-learning algorithms to identify clusters where HTE may be detected. Methods: Five unsupervised: Latent class analysis (LCA), K-means, partition around medoids, hierarchical, and spectral clustering; and four supervised algorithms: model-based recursive partitioning, Causal Forest (CF), and X-learner with Random Forest (XL-RF) and Bayesian Additive Regression Trees were individually applied to three prior ARDS RCTs. Clinical data and research protein biomarkers were used as partitioning variables, with the latter excluded for secondary analyses. For a clustering schema, HTE was evaluated based on the interaction term of treatment group and cluster with day-90 mortality as the dependent variable. Findings: No single algorithm identified clusters with significant HTE in all three trials. LCA, XL-RF, and CF identified HTE most frequently (2/3 RCTs). Important partitioning variables in the unsupervised approaches were consistent across algorithms and RCTs. In supervised models, important partitioning variables varied between algorithms and across RCTs. In algorithms where clusters demonstrated HTE in the same trial, patients frequently interchanged clusters from treatment-benefit to treatment-harm clusters across algorithms. LCA aside, results from all other algorithms were subject to significant alteration in cluster composition and HTE with random seed change. Removing research biomarkers as partitioning variables greatly reduced the chances of detecting HTE across all algorithms. Interpretation: Machine-learning algorithms were inconsistent in their abilities to identify clusters with significant HTE. Protein biomarkers were essential in identifying clusters with HTE. Investigations using machine-learning approaches to identify clusters to seek HTE require cautious interpretation. Funding: NIGMS R35 GM142992 (PS), NHLBI R35 HL140026 (CSC); NIGMS R01 GM123193, Department of Defense W81XWH-21-1-0009, NIA R21 AG068720, NIDA R01 DA051464 (MMC)

Today, the problem of climate change is addressed in many ways. Climate change would have a major impact on the hydrological cycle and consequently on available water resources. Therefore, the location and timing of the water resources in the world change. Water is an indispensable resource that connects all living things together and directly affects their lives. Water is a biological requirement not only for man, but also for economic, social, and cultural life itself. This study aims to identify similar countries in terms of water resources by taking into account the precipitation amounts together with climate change. For this purpose, time series clustering was conducted to precipitation amounts of 22 countries between 2000 and 2017 with the Hierarchical clustering and Partitioning Around Medoids (PAM) clustering methods based on two distance measures. Countries are divided into two clusters and they were evaluated according to their water risk levels and populations.

The COVID-19 pandemic has a strong worldwide impact on not only the health and economic sectors but also the (socio-)psychological functioning of individuals. Since psychological health is an important protective factor to prevent diseases, it is crucial to identify individuals with increased vulnerability during the crisis. 275 adults participated in a German online survey from April until August 2020 which investigated health-related, social, behavioral, and psychological effects of the COVID-19 pandemic. We here introduce an unsupervised clustering approach suitable for mixed data types combining the Gower distance with the Partitioning Around Medoids (PAM) algorithm k-Medoids. We were able to identify three clusters differing significantly in subjects’ well-being, psychological distress, and current financial and occupational concerns. The clusters also differed in age with younger persons reporting greater financial and occupational concerns, increased anxiety, higher psychological distress, and reduced subjective well-being. Features with the strongest impact on the clustering were examined using a wrapping method and the feature importance implemented in the random forest. Particularly, answers regarding financial and occupational concern, psychological distress, and current well-being were decisive for the assignment to a cluster. In summation, the clustering approach can identify persons with weakened psychological protective factors allowing them to provide tailored recommendations for preventive actions based on the cluster affiliation, e.g., via a web application.

Wireless Sensor Networks are affected by multipath fading, noise, and interference in the deployment environment, which makes wireless communication unreliable. Accurate and effective evaluation of communication link quality is helpful to assist the design of high-level routing protocol, ensure network throughput, and reduce retransmission. In this paper, a link quality estimation method based on ladder network is proposed: PAM(partitioning around medoids) algorithm is used to divide the link quality Grade, which as a basis for measuring the quality of the link; To improve the robustness and correlation of link features, ladder network is used to build link quality estimator to evaluate link quality. Data are collected in different interference scenarios. In the experiment, compared with other methods, the proposed link quality estimator has better precision and fl-score.

Consensus clustering algorithm, which integrates several clustering results obtained by common algorithms, can find a better result that is independent on parameter settings. However, this kind of algorithm is often designed based on simple, such as $K$ -means, algorithms, which is limited by the time complexity. In this work, a P system, a novel branch of bio-inspired computing with inherent parallel and distributed computation, is combined with the consensus clustering algorithm. As a result, an improved consensus clustering algorithm is constructed using the hierarchical membrane structure and parallel evolution mechanism in a cell-like P system with multi-catalysts, where the catalysts are utilized to regulate the parallelism of objects evolution. The integration strategy of the algorithm is based on a revised PAM where only the $q$ -nearest neighbors of the original medoids are considered as candidates for the new medoids. The experimental results indicate that the clustering quality of the proposed algorithm is more robust than well-known consensus clustering algorithms on data sets with noises and outliers. This work gives evidence that the effectiveness and efficiency of consensus clustering algorithms can be improved using P systems.

The increasing usage of the Internet and social networks has produced a significant amount of online textual data. These online textual data led to information overload and redundancy. It is important to eliminate the information redundancy and preserve the time required for reading these online textual data. Thus, there is a persistent need for an automatic text summarization system, which extract the relevant and salient information from a collection of documents, that sharing the same or related topics. Then, presenting this extracted information in a condensed form to preserve the main topics. This paper proposes an automatic, generic, and extractive Arabic multi-document summarization system. The proposed system employs the clustering-based and evolutionary multi-objective optimization methods. The clustering-based method discovers the main topics in the text, while the evolutionary multi-objective optimization method optimizes three objectives based on coverage, diversity/redundancy, and relevancy. The performance of the proposed system is evaluated using TAC 2011 and DUC 2002 datasets. The experimental results are compared using ROUGE evaluation measure. The obtained results showed the effectiveness of the proposed system compared to other peer systems. The proposed system outperformed other peer systems for all ROUGE metrics using TAC 2011. We achieved an F-measure of 38.9%, 17.7%, 35.4%, and 15.8% for Rouge-1, Rouge-2, Rouge-L, and Rouge-SU4, respectively. In addition, the proposed system with DUC 2002 dataset achieved an F-measure of 47.1%, 23.7%, 47.1%, 20.4% for Rouge-1, Rouge-2, Rouge-L, and Rouge-SU4, respectively.

In a centralized cooperative spectrum sensing (CSS) system, it is vulnerable to malicious users (MUs) sending fraudulent sensing data, which can severely degrade the performance of CSS system. To solve this problem, we propose sensing data fusion schemes based on K-medoids and Mean-shift clustering algorithms to resist the MUs sending fraudulent sensing data in this paper. The cognitive users (CUs) send their local energy vector (EVs) to the fusion center which fuses these EVs as an EV with robustness by the proposed data fusion method. Specifically, this method takes a Medoids of all EVs as an initial value and searches for a high-density EV by iteratively as a representative statistical feature which is robust to malicious EVs from MUs. It does not need to distinguish MUs from CUs in the whole CSS process and considers constraints imposed by the CSS system such as the lack of information of PU and the number of MUs. Furthermore, we propose a global decision framework based on fast K-medoids or Mean-shift clustering algorithm, which is unaware of the distributions of primary user (PU) signal and environment noise. It is worth noting that this framework can avoid the derivation of threshold. The simulation results reflect the robustness of our proposed CSS scheme.

El tema de la mediterraneidad ha generado bastante interes en el debate geopolitico, siendo Bolivia uno de los actores principales. Este hecho, junto con las nuevas herramientas de analisis de datos, como la inteligencia artificial y la mineria de datos, motivan el presente este estudio, el cual es pionero dentro de la literatura en el marco del analisis de paises sin salida maritima mediante algoritmos no supervisados de mineria de datos. En este sentido, se estudia y aplica la teoria de formacion de clusters a traves de los algoritmos K-means y PAM (Partitioning Around Medoids) con informacion de indicadores de comercio internacional de 188 paises de un periodo de diez anos, con el proposito de detectar si la condicion de mediterraneidad es un factor limitante en la dinamica comercial de los paises. Los resultados muestran que un subconjunto reducido de los paises mediterraneos, entre ellos Bolivia, habrian aliviado, durante la ultima decada, las restricciones que la mediterraneidad implica en los costos y tiempos de exportacion e importacion.

Globular Clusters (GCs) have historically been subdivided in either two (disk/halo) or three (disk/inner-halo/outer-halo) groups based on their orbital, chemical and internal physical properties. The qualitative nature of this subdivision makes it impossible to determine whether the natural number of groups is actually two, three, or more. In this paper we use cluster analysis on the $(\log M, \log \sigma_0, \log R_e, [Fe/H], \log | Z |)$ space to show that the intrinsic number of GC groups is actually either $k=2$ or $k=3$, with the latter being favored albeit non-significantly. In the $k=2$ case, the Partitioning Around Medoids (PAM) clustering algorithm recovers a metal-poor halo GC group and a metal-rich disk GC group. With $k=3$ the three groups can be interpreted as disk/inner-halo/outer-halo families. For each group we obtain a medoid, i.e. a representative element (NGC $6352$, NGC $5986$, and NGC $5466$ for the disk, inner halo, and outer halo respectively), and a measure of how strongly each GC is associated to its group, the so-called silhouette width. Using the latter, we find a correlation with age for both disk and outer halo GCs where the stronger the association of a GC with the disk (outer halo) group, the younger (older) it is., Comment: 17 figures, MNRAS submitted

Telematics car driving data describes drivers’ driving characteristics. This paper studies the driving characteristics at different speeds and their predictive power for claims frequency modeling. We first extract covariates from telematics car driving data using K -medoids clustering and principal components analysis. These telematics covariates are then used as explanatory variables for claims frequency modeling, in which we analyze their predictive power. Moreover, we use these telematics covariates to challenge the classical covariates usually used in practice.

Preparation for the modernization of the Kedung Putri Irrigation System (DI Kedung Putri) required a comprehensive assessment of the irrigation pillars, one of which was at the secondary level. To facilitate the assessment and development plan, a clustering was carried out using the k-medoids method, that used a representative data (called medoid) as the cluster center. Then, the decision making was conducted by using the Analytic Hierarchy Process (AHP) method. Performance assessment of 21 secondary channels was stated as the readiness index of irrigation modernization (IKMI). The assessment result showed that 9,52% included in good criteria, 71,43% included in fair criteria, and 19,05% included in poor criteria. Based on these results that DI Kedung Putri was not ready yet to be modernized. For this reason, it was necessary to conduct the system improvement in groups, namely by grouping based on similarities (clustering). The used method was k-medoids clustering using Rapid Miner 9.0 software. The clustering result showed that the optimal cluster number were 4 clusters, with the Davies Bouldin Index (DBI) value -1,959. The members of the 0, 1, 2 and 3 cluster were 6, 6, 8 and 1 secondary channels, respectively. Furthermore, the priority scale in clusters development was needed based on the performance of irrigation pillars on secondary channels. The results of AHP analysis showed that the order of priority development starts from cluster 0, followed by cluster 2, 1, and 3. The recommendations for the development of secondary channels incorporated in cluster, such as increasing water supply, routine infrastructure maintenance, technical assistance, and public campaigns in irrigation management. The secondary channel incorporated in cluster 3 had good performance on all pillars, so it only needed to maintain the existing operation and maintenance patterns., Preparation for the modernization of the Kedung Putri Irrigation System (DI Kedung Putri) required a comprehensive assessment of the irrigation pillars, one of which was at the secondary level. To facilitate the assessment and development plan, a clustering was carried out using the k-medoids method, that used a representative data (called medoid) as the cluster center. Then, decision making was conducted by using the Analytic Hierarchy Process (AHP) method. The performance assessment of 21 secondary channels was stated as the readiness index of irrigation modernization (IKMI). The assessment result showed that 9.52% belonged to good criteria, 71.43% belonged to fair criteria, and 19.05% belonged to poor criteria. Based on these results that DI Kedung Putri was considered not ready yet to be modernized. For this reason, it was necessary to conduct the system improvement in groups, namely by grouping based on similarities (clustering). The used method was k-medoids clustering using Rapid Miner 9.0 software. The clustering result showed that the optimal cluster number was 4clusters, with the Davies Bouldin Index (DBI) value -1.959. The members of the 0, 1, 2 and 3 clusters were 6, 6, 8 and 1 secondary channels, respectively. Furthermore, a priority scale in clusters development was needed based on the performance of irrigation pillars on secondary channels. The results of AHP analysis showed that the order of priority development starts from cluster 0, followed by cluster 2, 1 and 3. The recommendations for the development of secondary channels incorporated in cluster 0, such as increasing water supply, routine infrastructure maintenance, technical assistance, and public campaigns in irrigation management. The secondary channels incorporated in cluster 3 had good performance on all pillars, so it only needed to maintain the existingoperation and maintenance patterns.

A nondestructive multiparametric protocol for chert investigations (NM-PCI) was developed to provide objective criteria for raw material procurements and provenance investigations of archaeological cherts. For this purpose, macroscopic and mesoscopic characteristics (structures, textures and paleontological contents) were used to describe the main visible chert features as well as colourimetric and geochemical data acquired with portable devices. Data handling was performed by means of cluster analysis multivariate statistics, with an appropriate selection of algorithms to process and visualize mixed data matrices (partitioning around medoids (PAM), t-distributed stochastic neighbor embedding (t-SNE)). The protocol was applied to chert samples from the Gargano Promontory (Italy, Apulia) mining complex located in the districts of Peschici, Vieste and Mattinata. These mines, exploited for approximately four millennia, constitute an exceptional prehistoric record of mining archaeology and, above all, a key tool to understand the Neolithisation process throughout the Adriatic region. A map of the primary and secondary chert sources of the Gargano Promontory was produced to manage field sampling of mine structures. Waste materials were selected for the geologic samples of the five mines. Due to their inaccessibility, the samples from two mines consisted exclusively of waste materials. The NM-PCI proved able to discriminate most of the analysed mines and is well suited to provenance investigations of archaeological lithic industries. Compared to previous geochemical studies, the nondestructive method proposed here improves geological resolution, allowing the discrimination of cherts within the same formation.

Integration of multi-omics in cardiovascular diseases (CVDs) presents high potentials for translational discoveries. By analyzing abundance levels of heterogeneous molecules over time, we may uncover biological interactions and networks that were previously unidentifiable. However, to effectively perform integrative analysis of temporal multi-omics, computational methods must account for the heterogeneity and complexity in the data. To this end, we performed unsupervised classification of proteins and metabolites in mice during cardiac remodeling using two innovative deep learning (DL) approaches. First, long short-term memory (LSTM)-based variational autoencoder (LSTM-VAE) was trained on time-series numeric data. The low-dimensional embeddings extracted from LSTM-VAE were then used for clustering. Second, deep convolutional embedded clustering (DCEC) was applied on images of temporal trends. Instead of a two-step procedure, DCEC performes a joint optimization for image reconstruction and cluster assignment. Additionally, we performed K-means clustering, partitioning around medoids (PAM), and hierarchical clustering. Pathway enrichment analysis using the Reactome knowledgebase demonstrated that DL methods yielded higher numbers of significant biological pathways than conventional clustering algorithms. In particular, DCEC resulted in the highest number of enriched pathways, suggesting the strength of its unified framework based on visual similarities. Overall, unsupervised DL is shown to be a promising analytical approach for integrative analysis of temporal multi-omics.

Globalization trend in integrated circuit design and manufacturing process has increased the vulnerability of integrated circuit. These vulnerabilities mainly caused by hardware Trojan have a serious impact on the security of integrated circuits. Although side-channel analysis approach is the most promising Trojan detection approach, nearly all side-channel analysis approaches rely heavily on the availability of golden chips, which are extremely difficult to obtain. In this paper, a golden layout model instead of fabricated golden chips is introduced for the practical application of hardware Trojan detection approaches. The simulated voltage variations generated from the golden layout model at different process corners serve as golden reference, thus fabricated golden chips are not required during detection. Further, silicon measurements are performed to obtain the voltage variations of fabricated chips, and a model calibration algorithm is utilized to calibrate the golden model in the presence of process variations and random noise. Finally, the Trojan detection is formulated as a two-class classification problem, and the Trojan is identified using the partitioning around medoids algorithm. Experimental results demonstrate that the similarities between the simulated traces and measured traces are greater than 98.81%, and the proposed approach distinguishes the Trojan chips correctly even under ± 15% process variation.

In finance, cluster analysis is a tool particularly useful for classifying stock market multivariate time series data related to daily returns, volatility daily stocks returns, commodity prices, volume trading, index, enhanced index tracking portfolio, and so on. In the literature, following different methodological approaches, several clustering methods have been proposed for clustering multivariate time series. In this paper by adopting a fuzzy approach and using the Partitioning Around Medoids strategy, we suggest to cluster multivariate financial time series by considering the dynamic time warping distance. In particular, we proposed a robust clustering method capable to neutralize the negative effects of possible outliers in the clustering process. The clustering method achieves its robustness by adopting a suitable trimming procedure to identify multivariate financial time series more distant from the bulk of data. The proposed clustering method is applied to the stocks composing the FTSE MIB index to identify common time patterns and possible outliers.

In our work, we cluster the data collected on pregnancy women through hospitals, into three different categories such as elective LSCS, emergency LSCS and normal delivery. We aim to identify outliers so that we can predict the complications that can occur during pregnancy. These complications can even lead to maternal death. Clustering is performed by the aid of k-means, and k-medoid procedures which are implemented in windows form application to run in Microsoft Visual Studio, a comparison is provided on its performance and efficiency between both the algorithms and also the outlier, i.e. normal delivery cases are detected and depicted. From our work, we can clearly say that K-Means works better when the dataset is between the range of 50 – 100. Suppose the data set is greater than 500 then K-medoid works efficiently. In our work, we are considering various maternity cases. Here we are detecting normal delivery cases as an outlier.

This paper presents a data mining study and cluster analysis of social data obtained on small producers and family farmers from six country cities in Ceará state, northeast Brazil. The analyzed data involve demographic, economic, agriculture and food insecurity information. The goal of the study is to establish profiles for the small producer families that reside in the region and to identify relevant features which differentiate these profiles. Moreover, we provide an efficient data mining methodology for analysis of social data sets which is capable of handling its natural challenges, such as mixed variables and abundance of null values. We use the Silhouette method for the estimation of the best number of natural groups within the data, along with the Partitioning Around Medoids clustering algorithm in order to compute the profiles. The Correlation-Based Feature Selection method is used to identify which social criteria are the most important to differentiate the families from each profile. Classification models based on support vector machines, multilayer perceptron and decision trees were developed aiming to predict in which of the identified clusters an arbitrary family would be best fit. We obtained a good separation of the families into two clusters, and a multilayer perceptron model with approximately 93.5% prediction accuracy. Keywords: Clustering, Social data, Classification, Pam, Data mining

This study investigates the use of clustering methodologies as a means of reducing spatio-temporal wind speed data into statistically representative classes of temporal profiles for further processing and interpretation. The clustering methodologies are applied to the high-resolution spatio-temporal, meso-scale renewable energy resource dataset produced for Southern Africa by the Council of Scientific and Industrial Research. This large dataset incorporates thousands of coordinates and represents a challenge from a computational perspective. This dataset can be reduced by applying clustering techniques to classify the temporal wind speed profiles into categories with similar statistical properties. Various clustering algorithms are considered, with the view to compare the performances of these algorithms for large wind resource datasets, namely k-means, partitioning around medoids, the clustering large applications algorithm, agglomerative clustering, the divisive analysis algorithm and fuzzy c-means clustering. Two distance measures are considered, namely the Euclidean distance and Pearson correlation distance. The validation metrics evaluated in the investigation includes the silhouette coefficient, the Calinski-Harabasz index and the Dunn index. Case study results are presented for the Komsberg Renewable Energy Development Zone, located in Western Cape, South Africa. This zone is selected based on the high mean wind speed and large standard deviation exhibited by the temporal wind speed profiles associated with the zone. The effects of seasonal variation in the temporal wind speed profiles are considered by partitioning the input dataset in accordance with the low and high demand seasons defined by the Megaflex Time of Use tariff. The clustered wind resource maps produced by the proposed methodology represent a valuable input dataset for further studies such as siting and the optimal geographical allocation of wind generation capacity to reduce the variability and ramping effects that are inherent to wind energy.

The k-medoids method is a non-hierarchical clustering to classify n object into k clusters that have the same characteristics. This clustering algorithm uses the medoid as its cluster center. Medoid is the most centrally located object in a cluster, so it’s robust to outliers. In cluster analysis the objects are grouped by the similarity. To measure the similarity, it can be used distance measures, euclidean distance and cityblock distance. The distance that is used in cluster analysis can affect the clustering results. Then, to determine the quality of the clustering results can be used the internal criteria with silhouette width and C-index. In this research the k-medoids method to classify of regencies/cities in Central Java based on type and number of crimes. The optimal cluster at k= 4 use euclidean distance, where the silhouette index= 0,3862593 and C-index= 0,043893. Keywords: Clustering, k-Medoids, Euclidean distance, Cityblock distance, Silhouette index, C-index, Crime

The process of clustering in the general perspective is limited to the grouping of data into clusters and finds its applications in the fields of information retrieval, text ranking and classification and more. The dimension of e-Learning is to improve learning with various tools and technologies. Grouping of learners based on their learning levels is found to improve the learning abilities. Scientific method to cluster the learners is not available in literature, which can further simplify the amalgamation of learning complemented through clustering. This paper is an attempt to examine the aspects of implementing clustering to group the learners according to their learning abilities.

High dimensional data analysis within relational database management systems (RDBMS) is challenging because of inadequate support from SQL. Currently, subspace clustering of high dimensional data is implemented either outside DBMS using wrapper code or inside DBMS using SQL User Defined Functions/Aggregates(UDFs/UDAs). However, both these approaches have potential disadvantages from performance, resource usage, and security perspective for voluminous and frequently updated data. Hence, we propose an efficient querying system, named SubspaceDB, that implements subspace clustering directly within an RDBMS. SubspaceDB provides a novel set of query operators, each with an optimization objective, to facilitate interactive analysis for subspace clustering. The query operators focus on retrieving optimal answers to four key query types : (a) Medoid queries, (b) Neighbourhood queries, (c) Partial similarity queries, and (d) Prominence queries, that aid the formation of subspace clusters. Experimental studies on real and synthetic databases of size 15 M tuples and 104 attributes show that our proposed approach SubspaceDB can be over 10 times faster as compared to a conventional wrapper-based or SQL UDF approach. The proposed approach is also efficient in retrieving at least 50% data with performance improvement of at least 25%.

The definition of management zones (MZs) in agricultural fields has been suggested as an economically viable approach to precision agriculture. The most used methods for this task are the cluster analysis algorithms Fuzzy C-means (FCM) and K-means. However, some studies have presented that these algorithms may not provide the best classes to define MZs. Considering that these studies presented comparisons of only a few clustering methods, the objective of our research was to evaluate 20 algorithms for defining MZs, including more than 10 methods that have not been investigated in the literature for this purpose. The following algorithms were evaluated: Average Linkage, Bagged Clustering, Centroid Linkage, Clara, Complete Linkage, Diana, Fanny, FCM, Fuzzy C-shells, Hard Competitive Learning, Hybrid Hierarchical Clustering, K-means, McQuitty's Method, Median Linkage, Neural Gas, Partitioning Around Medoids, Single Linkage, Spherical K-means, Unsupervised Fuzzy Competitive Learning and Ward's Method. The evaluation was conducted with data obtained between 2010 and 2015 from three commercial agricultural fields cultivated with soya bean and maize in Brazil. The results of the analysis of variance suggested a division of the three fields into two classes with significantly different yields and a division of one of the fields into three classes. These divisions were satisfactorily performed using 17 algorithms, but McQuitty's Method and Fanny were considered to be the best choices because they produced the largest reductions in the variance of the yield in the three fields. In addition, they generated classes with high internal homogeneity and delimited MZs without spatial fragmentation.

Clustering geographical units based on a set of quantitative features observed at several time occasions requires to deal with the complexity of both space and time information. In particular, one should consider (1) the spatial nature of the units to be clustered, (2) the characteristics of the space of multivariate time trajectories, and (3) the uncertainty related to the assignment of a geographical unit to a given cluster on the basis of the above complex features. This paper discusses a novel spatially constrained multivariate time series clustering for units characterized by different levels of spatial proximity. In particular, the Fuzzy Partitioning Around Medoids algorithm with Dynamic Time Warping dissimilarity measure and spatial penalization terms is applied to classify multivariate Spatial–Temporal series. The clustering method has been theoretically presented and discussed using both simulated and real data, highlighting its main features. In particular, the capability of embedding different levels of proximity among units, and the ability of considering time series with different length.

The recurrent question about the effectiveness of agri-environmental measure (AEM) in Sardinia (Italy) is whether European Union (EU) funds allocate resources to where they are most needed. To answer this question, a spatial approach is suggested, namely an approach that considers geography as a factor in measuring the success of such policy. A geographical approach can be used to pinpoint “hotspots” in order to determine an appropriate distribution of funds. To implement such an approach to the distribution of EU funding, a Spatial Fuzzy Partitioning Around Medoids (SFPAM) analysis is advocated. The contribution of this research is that it combines a temporal dimension within an explicitly spatial approach. It achieves this by using a dataset that includes both geographical and economic factors such as farm sizes, their management, the number of organic farms involved, the agriculture area invested by the AEM and the size of the workforce involved. Its strategy is the identification of medoids which are represented by a specific municipality. This allows the identification of aggregated neighborhoods for the visualization of AEM outcomes based on a fuzzy partitioning method. The results provide useful policy implications to determine where and when financial efforts should be renewed, where to negotiate sustainable development strategies, and how to expand spatially the benefits of financial funding to other agricultural measures, such as technological innovations in agriculture, reforestation programs, marketing strategies, climate change mitigation, and rural development.

Modeling time-varying operations in complex energy systems optimization problems is often computationally intractable, and time-series input data are thus often aggregated to representative periods. In this work, we introduce a framework for using clustering methods for this purpose, and we compare both conventionally-used methods (k-means, k-medoids, and hierarchical clustering), and shape-based clustering methods (dynamic time warping barycenter averaging and k-shape). We compare these methods in the domain of the objective function of two example operational optimization problems: battery charge/discharge optimization and gas turbine scheduling, which exhibit characteristics of complex optimization problems. We show that centroid-based clustering methods represent the operational part of the optimization problem more predictably than medoid-based approaches but are biased in objective function estimate. On certain problems that exploit intra-daily variability, such as battery scheduling, we show that k-shape improves performance significantly over conventionally-used clustering methods. Comparing all locally-converged solutions of the clustering methods, we show that a better representation in terms of clustering measure is not necessarily better in terms of objective function value of the optimization problem.

Induction motor is widely used in industrial applications. The research paper presents the diagnosis of the induction motor bearing faults using Discrete Wavelet Transforms and Partitioning Around Medoids algorithm methods. The experimental test rig was developed to obtain data of the bearings on healthy or damaged conditions. Several mother-level wavelets are tried in order to get the best performance to find bearing faults. The wavelet transform results are used as an input of the Partitioning Around Medoids algorithm to cluster the bearing condition. The results showed that the methods proposed could provide an accurate diagnosis of the bearing condition.

BACKGROUND Behavioral representations obtained from mobile sensing data can be helpful for the prediction of an oncoming psychotic relapse in patients with schizophrenia and the delivery of timely interventions to mitigate such relapse. OBJECTIVE In this study, we aim to develop clustering models to obtain behavioral representations from continuous multimodal mobile sensing data for relapse prediction tasks. The identified clusters can represent different routine behavioral trends related to daily living of patients and atypical behavioral trends associated with impending relapse. METHODS We used the mobile sensing data obtained from the CrossCheck project for our analysis. Continuous data from six different mobile sensing-based modalities (ambient light, sound, conversation, acceleration, etc) obtained from 63 patients with schizophrenia, each monitored for up to a year, were used for the clustering models and relapse prediction evaluation. Two clustering models, Gaussian mixture model (GMM) and partition around medoids (PAM), were used to obtain behavioral representations from the mobile sensing data. These models have different notions of similarity between behaviors as represented by the mobile sensing data, and thus, provide different behavioral characterizations. The features obtained from the clustering models were used to train and evaluate a personalized relapse prediction model using balanced random forest. The personalization was performed by identifying optimal features for a given patient based on a personalization subset consisting of other patients of similar age. RESULTS The clusters identified using the GMM and PAM models were found to represent different behavioral patterns (such as clusters representing sedentary days, active days but with low communication, etc). Although GMM-based models better characterized routine behaviors by discovering dense clusters with low cluster spread, some other identified clusters had a larger cluster spread, likely indicating heterogeneous behavioral characterizations. On the other hand, PAM model-based clusters had lower variability of cluster spread, indicating more homogeneous behavioral characterization in the obtained clusters. Significant changes near the relapse periods were observed in the obtained behavioral representation features from the clustering models. The clustering model-based features, together with other features characterizing the mobile sensing data, resulted in an F2 score of 0.23 for the relapse prediction task in a leave-one-patient-out evaluation setting. The obtained F2 score was significantly higher than that of a random classification baseline with an average F2 score of 0.042. CONCLUSIONS Mobile sensing can capture behavioral trends using different sensing modalities. Clustering of the daily mobile sensing data may help discover routine and atypical behavioral trends. In this study, we used GMM-based and PAM-based cluster models to obtain behavioral trends in patients with schizophrenia. The features derived from the cluster models were found to be predictive for detecting an oncoming psychotic relapse. Such relapse prediction models can be helpful in enabling timely interventions.

The variability in generation introduced in the electrical system by an increasing share of renewable technologies must be addressed by balancing mechanisms, demand response being a prominent one. In parallel, the massive introduction of smart meters allows for the use of high frequency energy use time series data to segment electricity customers according to their demand response potential. This paper proposes a smart meter time series clustering methodology based on a two-stage k-medoids clustering of normalized load-shape time series organized around the day divided into 48 time points. Time complexity is drastically reduced by first applying the k-medoids on each customer separately, and second on the total set of customer representatives. Further time complexity reduction is achieved using time series representation with low computational needs. Customer segmentation is undertaken with only four easy-to-interpret features: average energy use, energy–temperature correlation, entropy of the load-shape representative vector, and distance to wind generation patterns. This last feature is computed using the dynamic time warping distance between load and expected wind generation shape representative medoids. The two-stage clustering proves to be computationally effective, scalable and performant according to both internal validity metrics, based on average silhouette, and external validation, based on the ground truth embedded in customer surveys.

International audience; This paper deals with supervised classification and feature selection with application in the context of high dimensional features. A classical approach leads to an optimization problem minimizing the within sum of squares in the clusters (2 norm) with an 1 penalty in order to promote sparsity. It has been known for decades that 1 norm is more robust than 2 norm to outliers. In this paper, we deal with this issue using a new proximal splitting method for the minimization of a criterion using 1 norm both for the constraint and the loss function. Since the 1 criterion is only convex and not gradient Lipschitz, we advocate the use of a Douglas-Rachford minimization solution. We take advantage of the particular form of the cost and, using a change of variable, we provide a new efficient tailored primal Douglas-Rachford splitting algorithm which is very effective on high dimensional dataset. We also provide an efficient classifier in the projected space based on medoid modeling. Experiments on two biological datasets and a computer vision dataset show that our method significantly improves the results compared to those obtained using a quadratic loss function.; Ce document traite de la classification supervisée et de la sélection des descripteurs avec application dans le contexte des descripteurs de haute dimension. Une approche classique conduit à un problème d'optimisation minimisant la norme quadratique (norme 2) avec une pénalite de norme 1 afin de promouvoir la parcimonie. Il est bien connu que la norme 1 est plus robuste que la norme 2 aux valeurs aberrantes. Dans cet article, nous abordons cette question avec une nouvelle méthode proximale pour la minimisation d'un critère utilisant la norme 1 à la fois pour la contrainte et la fonction de perte. Comme le critère 1 n'est que convexe et non gradient de Lipschitz, nous proposons l'utilisation d'une solution de minimisation de type Douglas-Rachford. Nous tirons parti de la forme du critère et, en utilisant un changement de variable, nous fournissons un nouvel algorithme de type Douglas-Rachford dans le primal qui est très efficace dans le contexte des descripteurs de haute dimension. Nous fournissons également un classifieur efficace dans l'espace projeté basé sur la modélisation de medoïds. Des expériences sur des données biologiques et de vision par ordinateur montrent que notre méthode améliore les résultats par rapport à ceux obtenus avec une fonction de perte quadratique.

The recent rise of big data analytics is transforming the apparel retailing industry. E-retailers, for example, effectively use large volumes of data generated as a result of their day-to-day business operations data to aid operations and supply chain management. Although logs of how consumers navigate through an e-commerce website are readily available in a form of clickstream data, clickstream analysis is rarely used to derive insights that can support marketing decisions, leaving it an under-researched area of study. Adding to this research stream by exploring the case of a U.K.-based fast-fashion retailer, this article reveals unique consumer segments and links them to the revenue they are capable of generating. Applying the partitioning around medoids algorithm to three random samples of 10 000 unique consumer visits to the e-commerce site of a fast-fashion retailer, six consumer segments are identified. This article shows that although the “mobile window shoppers” segment consists of the largest consumer segment, it attracts the lowest revenue. In contrast, “visitors with a purpose,” although one of the smallest segments, generates the highest revenue. The findings of this article contribute to marketing research and inform practice, which can use these insights to target customer segments in a more tailored fashion.

More and more data are recorded continuously in an interval or discretely only at finite time points nowadays due to the frequency of data collection. These data are functional data which belong to big data. This paper introduces functional clustering in functional data analysis. We propose a weighted partitioning around medoid algorithm with estimation of number of clusters for functional clustering, and use an example to prove its effectiveness.

COVID-19 is a viral infection that affects people differently, where the majority of cases develop mild symptoms, some people require hospitalization, and unfortunately, a small number of patients perish. Hence, identifying risk factors is critical for physicians to make treatment decisions. The purpose of this article is to determine whether unsupervised analysis of risk factors in positive and negative COVID-19 subjects can aid in the identification of a set of reliable and clinically relevant risk profiles. Positive and negative patients hospitalized were randomly selected from the Mexican Open Registry between March and May 2020. Thirteen risk factors, three distinct outcomes, and COVID-19 test results were used to categorize registry patients. As a result, the dataset was reported via 6144 different risk profiles for each age group. The unsupervised learning method is proposed in this study to discover the most prevalent risk profiles. The data was partitioned into discovery (70%) and validation (30%) sets. The discovery set was analyzed using the partition around medoids (PAM) method, and the stable set of risk profiles was estimated using robust consensus clustering. The PAM models' reliability was validated by predicting the risk profile of subjects from the validation set and patients admitted in November 2020. In the validation set, the clinical relevance of the risk profiles was evaluated by determining the prevalence of three patient outcomes: pneumonia diagnosis, ICU admission, or death. Six positive and five negative COVID-19 risk profiles were identified, with significant statistical differences between them. As a result, PAM clustering with consensus mapping is a viable method for discovering unsupervised risk profiles in subjects with severe respiratory health problems.

Recently, learning only from ordinal information of the type “itemxis closer to itemythan to itemz” has received increasing attention in the machine learning community. Such triplet comparisons are particularly well suited for learning from crowdsourced human intelligence tasks, in which workers make statements about the relative distances in a triplet of items. In this paper, we systematically investigate comparison-based centrality measures on triplets and theoretically analyze their underlying Euclidean notion of centrality. Two such measures already appear in the literature under opposing approaches, and we propose a third measure, which is a natural compromise between these two. We further discuss their relation to statistical depth functions, which comprise desirable properties for centrality measures, and conclude with experiments on real and synthetic datasets for medoid estimation and outlier detection.

Sentiment analysis explores the views, perceptions and feelings of people concerning entities like subjects, goods, organizations, resources and individuals. The opinion of some people in social network influences the opinion behavior and thoughts of other people. They are known as influential user. In this article, both the sentiment analysis and identification of influential user are proposed. Initially, Twitter data are preprocessed by proposing weighted partition around medoids (WPAM) with artificial cooperative search (WPAM-ACS) which extracts topics from Twitter data through dynamic clustering (DC). For sentiment classification, NLP has been used in many works. The main issue of using NLP for sentiment classification is that many languages do not have the adequate resources to develop NLP models. So, a fuzzy deep neural network (FDNN) is proposed in this paper for sentiment classification, because FDNN effectively handles the uncertainties and noises in tweet data than other state of the arts. Emotional conformity is a metric that refers to how people from an emotional point of view agree with another person. It is given as additional input to FDNN along with the tweets for sentiment classification. Finally, influential users are detected by temporal influential model (TIM) formulated as likelihood function using incremental logistic regression (ILLR) in which user’s opinion sequence is considered for identification of influential user. In the experimental results, sentiment analysis is evaluated in terms of precision, recall and F-measure and proved that the proposed DC–FDNN sentiment classification is better than fixed clustering and NLP (FC–NLP)-based sentiment classification. Influential user detection using TIM-ILLR on opinion sequences which are identified by DC-FDNN is evaluated in terms of accuracy and proved that TIM-ILLR is better than other methods such as maximum likelihood estimation (MLE), support vector regression (SVR) and logistic regression (LR).

The Classroom Observation Protocol for Undergraduate STEM (COPUS) provides descriptive feedback to instructors by capturing student and instructor behaviors occurring in the classroom. Due to the increasing prevalence of COPUS data collection, it is important to recognize how researchers determine whether groups of courses or instructors have unique classroom characteristics. One approach uses cluster analysis, highlighted by a recently developed tool, the COPUS Analyzer, that enables the characterization of COPUS data into one of seven clusters representing three groups of instructional styles (didactic, interactive, and student centered). Here, we examine a novel 250 course data set and present evidence that a predictive cluster analysis tool may not be appropriate for analyzing COPUS data. We perform a de novo cluster analysis and compare results with the COPUS Analyzer output and identify several contrasting outcomes regarding course characterizations. Additionally, we present two ensemble clustering algorithms: 1) k-means and 2) partitioning around medoids. Both ensemble algorithms categorize our classroom observation data into one of two clusters: traditional lecture or active learning. Finally, we discuss implications of these findings for education research studies that leverage COPUS data.

The current development of quantum mechanics and its applications suppose a threat to modern cryptography as it was conceived. The abilities of quantum computers for solving complex mathematical problems, as a strong computational novelty, is the root of that risk. However, quantum technologies can also prevent this threat by leveraging quantum methods to distribute keys. This field, called Quantum Key Distribution (QKD) is growing, although it still needs more physical basics to become a reality as popular as the Internet. This work proposes a novel methodology that leverages medoid-based clustering techniques to design quantum key distribution networks on commercial fiber optics systems. Our methodology focuses on the current limitations of these communication systems, their error loss and how trusted repeaters can lead to achieve a proper communication with the current technology. We adapt our model to the current data on a wide territory covering an area of almost 100,000 km 2 , and prove that considering physical limitations of around 45 km with 3.1 error loss, our design can provide service to the whole area. This technique is the first to extend the state of the art network’s design, that is focused on up to 10 nodes, to networks dealing with more than 200 nodes.

Capacity expansion planning (CEP) of power systems determines the optimal future generation mix and/or transmission lines. Due to the increasing penetration of renewables, CEP has to capture the hourly variations of renewable generator outputs and load demand. Since CEP problems typically involve planning horizons of several years, solving the fullspace models where the operating decisions corresponding to all the days is intractable. Therefore, some “representative days” are selected as a surrogate to the fullspace model. We present an input-based and a cost-based approach in combination with the k -means and the k -medoids clustering algorithms for representative day selection. The mathematical properties of the proposed algorithms are analyzed, including an approach to calculate the “optimality gap” of the investment decisions obtained from the representative day model to the fullspace model, and the relationship between the clustering error and the optimality gap. To capture the extreme operating conditions, two novel approaches, i.e., a “load shedding cost” approach and a “highest cost” approach, are proposed to identify the “extreme days”. We conclude with a case study based on the Electric Reliability Council of Texas (ERCOT) region, which compares the different approaches and the effects of adding the extreme days.

COVID-19 is a viral disease that affects people in different ways: Most people will develop mild symptoms; others will require hospitalization, and a few others will die. Hence identifying risk factors is vital to assist physicians in the treatment decision. The objective of this paper is to determine whether unsupervised analysis of risk factors of positive and negative COVID-19 subjects may be useful for the discovery of a small set of reliable and clinically relevant risk-profiles. We selected 13367 positive and 19958 negative hospitalized patients from the Mexican Open Registry. Registry patients were described by 13 risk factors, three different outcomes, and COVID-19 test results. Hence, the dataset could be described by 6144 different risk-profiles per age group. To discover the most common risk-profiles, we propose the use of unsupervised learning. The data was split into discovery (70%) and validation (30%) sets. The discovery set was analyzed using the partition around medoids (PAM) method and robust consensus clustering was used to estimate the stable set of risk-profiles. We validated the reliability of the PAM models by predicting the risk-profile of the validation set subjects. The clinical relevance of the risk-profiles was evaluated on the validation set by characterizing the prevalence of the three patient outcomes: pneumonia diagnosis, ICU, or death. The analysis discovered six positives and five negative COVID-19 risk-profiles with strong statistical differences among them. Henceforth PAM clustering with consensus mapping is a viable method for unsupervised risk-profile discovery among subjects with critical respiratory health issues.

A data-driven approach for the identification of local turbulent-flow states and of their dynamics is proposed. After subdividing a flow domain in smaller regions, the K -medoids clustering algorithm is used to learn from the data the different flow states and to identify the dynamics of the transition process. The clustering procedure is carried out on a two-dimensional (2-D) reduced-order space constructed by the multidimensional scaling (MDS) technique. The MDS technique is able to provide meaningful and compact information while reducing the dimensionality of the problem, and therefore the computational cost, without significantly altering the data structure in the state space. The dynamics of the state transitions is then described in terms of a transition probability matrix and a transition trajectory graph. The proposed method is applied to a direct numerical simulation dataset of an incompressible boundary layer flow developing on a flat plate. Streamwise-spanwise velocity fields at a specific wall-normal position are referred to as observations. Reducing the dimensionality of the problem allows us to construct a 2-D map, representative of the local turbulence intensity and of the spanwise skewness of the turbulence intensity in the observations. The clustering process classifies the regions containing streaks, turbulent spots, turbulence amplification and developed turbulence while the transition matrix and the transition trajectories correctly identify the states of the process of bypass transition. This work has been supported by: (i) the Madrid Government (Comunidad de Madrid) under the Multiannual Agreement with UC3M in the line of ‘Fostering Young Doctors Research’ (PITUFLOW-CM-UC3M), and in the context of the V PRICIT (Regional Programme of Research and Technological Innovation); (ii) the COTURB project (Coherent Structures in Wall-bounded Turbulence), funded by the European Research Council (ERC), under grant ERC-2014.AdG-669505.

This work studies clustering algorithms which operates with ordinal or comparison-based queries (operations), a situation that arises in many active-learning applications where “dissimilarities” between data points are evaluated by humans. Typically, exact answers are costly (or difficult to obtain in large amounts) while possibly erroneous answers have low cost. Motivated by these considerations, we study algorithms with non-trivial trade-offs between the number of exact (high-cost) operations and noisy (low-cost) operations with provable performance guarantees. Specifically, we study a class of polynomial-time graph-based clustering algorithms (termed Single-Linkage) which are widely used in practice and that guarantee exact solutions for stable instances in several clustering problems (these problems are NP-hard in the worst case). We provide several variants of these algorithms using ordinal operations and, in particular, non-trivial trade-offs between the number of high-cost and low-cost operations that are used. Our algorithms still guarantee exact solutions for stable instances of k-medoids clustering, and they use a rather small number of high-cost operations, without increasing the low-cost operations too much., Algorithms, 14 (2)

e-Learning is appropriate when the learners are grouped and facilitated to learn according to their learning style and at their own pace. Elaborate researches have been proposed to categorize learners based on various e-learning parameters. Most of these researches have deployed the clustering principles for grouping eLearners, and in particular, they have utilized K-Medoid principle for better clustering. In the classical K-Medoid algorithm, predicting or determining the value of K is critical, two methods namely the Elbow and Silhouette methods are widely applied. In this paper, we experiment with the application of both these methods to determine the value of K for clustering eLearners in K-Medoid and prove that Silhouette method best predicts the value of K.

International audience; Several sawmill simulators exist in the forest-product industry. They are able to simulate the sawing of a log to generate the set of lumbers that would be obtained by transforming a log at a sawmill. In particular, such simulators are able to use a 3D scan of the exterior shape of the logs as input for the simulation. However, it was observed that they can be computationally intensive. Therefore, several authors have proposed to use Artificial Intelligence metamodel, which, in general, can make predictions extremely fast once trained. Such models can approximate the results of a simulator using a vector of descriptive features representing a log, or, alternatively, the full 3D log scans. This paper proposes to use dissimilarity to representative log scans as features to train a Machine Learning classifier. The concept of class Medoids as representative elements of a class will be presented, and a Simlarity Discrimant Analysis was chosen as a good candidate ML classier. This classifier will be compared with two others models studied by the authors.

Cluster analysis is applied to a DNS dataset of a transitional boundary layer developing over a flat plate. The stream-wise-span-wise plane at a wall normal distance close to the wall is sampled at several time instants and discretized into small sub-regions, which are the observations analysed in this work. Using K-medoids clustering algorithm, a partition of the observations is sought such that the medoids in each cluster represent the main local states. The clustering has been carried out on a two-dimensional reduced-order feature space, constructed with the multi-dimensional scaling technique. The clustered feature space provides a partitioning which consists of five different regions. The observations are automatically classified as laminar, turbulent spots, amplification of disturbances, or fully-developed turbulence. The Lagrangian evolution of the regions and the state transitions are described as a Markov process in terms of transition probability matrix and transition trajectory graph to determine the transition dynamics between different states. PITUFLOW-CM-UC3M, funded by the call "Programa de apoyo a la realización de proyectos interdisciplinares de I+D para jóvenes investigadores de la Universidad Carlos III de Madrid 2019-2020" under the frame of the Convenio Plurianual Comunidad de Madrid-Universidad Carlos III de Madrid. COTURB, funded by the European Research Council, under grant ERC-2014-AdG-669505.

International audience; Predicting the set of lumbers which would be obtained from sawing a log at a specific sawmill is a difficult problem, which complicates short and mid term decision making in this industry. While sawmill simulators able to simulate the sawing of a log from a 3D scan of its outer shape exist, they can be extremely computationally intensive. Several alternative approaches based on machine learning algorithms and different set of features were explored in previous works. This paper proposes the use of one hidden layer perceptrons, and a vector of features build from dissimilarities from the scans to a set of selected wood logs, chosen as the class medoids. Several architectures are tested and compared to validate the pertinence of the proposed set of medoid-based features. The lowest mean squared error was obtained for MISO neural networks with a sigmoid output activation function, to constrain the output value ranges.

Data mining technique has been used to extract potentially useful knowledge from big data. However, data mining sometimes faces the issue of incorrect results which could be due to the presence of an outlier in the analyzed data. In the literature, it has been identified that the detection of this outlier could enhance the quality of the dataset. An important type of data that requires outlier detection for accurate prediction and enhanced decision making is time series data. Time series data are valuable as it helps to understand the past behavior which is helpful for future predictions hence, it is important to detect the presence of outliers in time series dataset. This paper proposes an algorithm for outlier detection in Multivariate Time Series (MTS) data based on a fusion of K-medoid, Standard Euclidean Distance (SED), and Z-score. Apart from SED, experiments were also performed on two other distance metrics which are City Block and Euclidean Distance. Z-score performance was compared to that of inter-quartile. However, the result obtained showed that the Z-score technique produced a better outlier detection result of 0.9978 F-measure as compared to inter-quartile of 0.8571 F-measure. Furthermore, SED performed better when combined with both Z-score and inter-quartile than City Block and Euclidean Distance.

Cluster analysis can be viewed as a result of the natural evolution of the vast amount of data from daily life, and can discover invisible feature information to contribute to the analysis. K-means algorithm is one of the wide data clustering methods in a variety of real-world applications thanks to its simpleness. However, the k-means is sensitive to noise and outlier data points because a small number of such data can substantially influence the mean value of the cluster. In light of this, the k-medoids algorithm selects a point as a new center that minimizes the sum of the dissimilarities in the cluster, to diminish such sensitivity to outliers. Nevertheless, the line of the k-medoids algorithm is limited by its amounts of computation and not to handle with data efficiently. To this end, we present a novel k-medoids algorithm motivated by the theory of ball cluster, relationship between clusters and partitioning cluster for assigning samples into their nearest medoids efficiently, called ball k-medoids, which drop the distance calculation of sample-medoid significantly. Moreover, a threshold is inferenced by the rollback method for reducing computation of medoid-medoid distance and accelerating clustering. Experiments finally demonstrate that the performance of ball k-medoids achieves more efficient in comparison with other k-medoids algorithms, and it performs exacter accuracy compared with k-means.