Clustering Algorithm For Detection Research Articles

832Detecting spatial clusters of human salmonellosis in Victoria

Abstract Focus of Presentation ‘Cluster Tracker’ is an automated tool for spatial cluster detection of notifiable disease data collected by the Department of Health (DH), Victoria. The tool combines R statistical software and a SaTScan cluster detection algorithm (prospective space-time permutation scan statistic) to detect notifiable disease case clusters in Victoria and is presently implemented for salmonellosis (categorised by type and/or MLVA). The objective of the tool is to conduct an initial screening of case data to improve the prioritisation of salmonellosis cases for epidemiological investigation. Findings The Cluster Tracker tool parameters have been validated using historical data from 2017-2018, comparing DH outbreak and cluster investigations identified by usual surveillance activities with clusters detected by the Cluster Tracker tool. Parameter selection considered cluster detection agreement and disagreement, disease-specific epidemiological characteristics, and operational requirements. The Cluster Tracker tool was able to provide closely-aligned agreement with existing DH outbreak and cluster investigations using the validated parameters. Implications This automated spatial cluster detection tool complements existing desktop surveillance of salmonellosis notifications to enhance public health decision making, and serves as an example of how spatial methods can improve real-time surveillance. Key messages Advanced spatial statistical tools have a role alongside traditional methods to make better use of limited epidemiological capacity and improve the timeliness and prioritisation of surveillance activities for notifiable diseases.

Evolutionary clustering and community detection algorithms for social media health surveillance

The prominent rise of social networks within the past decade have become a gold mine for data mining operations seeking to model the real world through these virtual worlds. One of the most important applications that has been proposed is utilizing information generated from social networks as a supplemental health surveillance system to monitor disease epidemics. At the time this research was conducted in 2020, the COVID-19 virus had evolved into a global pandemic, forcing many countries to implement preventative measures to halt its expanse. Health surveillance has been a powerful tool in placing further preventative measures, however it is not a perfect system, and slowly collected, misidentified information can prove detrimental to these efforts. This research proposes a new potential surveillance avenue through unsupervised machine learning using dynamic, evolutionary variants of clustering algorithms DBSCAN and the Louvain method to allow for community detection in temporal networks. This technique is paired with geographical data collected directly from the social media Twitter, to create an effective and accurate health surveillance system that grows as time passes. The experimental results show that the proposed system is promising and has the potential to be an advancement on current machine learning health surveillance techniques.

Socioeconomically Disadvantaged Neighborhoods Face Increased Persistence of SARS-CoV-2 Clusters.

Objective: To investigate the association between socioeconomic deprivation and the persistence of SARS-CoV-2 clusters.Methods: We analyzed 3,355 SARS-CoV-2 positive test results in the state of Geneva (Switzerland) from February 26 to April 30, 2020. We used a spatiotemporal cluster detection algorithm to monitor SARS-CoV-2 transmission dynamics and defined spatial cluster persistence as the time in days from emergence to disappearance. Using spatial cluster persistence measured outcome and a deprivation index based on neighborhood-level census socioeconomic data, stratified survival functions were estimated using the Kaplan-Meier estimator. Population density adjusted Cox proportional hazards (PH) regression models were then used to examine the association between neighborhood socioeconomic deprivation and persistence of SARS-CoV-2 clusters.Results: SARS-CoV-2 clusters persisted significantly longer in socioeconomically disadvantaged neighborhoods. In the Cox PH model, the standardized deprivation index was associated with an increased spatial cluster persistence (hazard ratio [HR], 1.43 [95% CI, 1.28–1.59]). The adjusted tercile-specific deprivation index HR was 1.82 [95% CI, 1.56–2.17].Conclusions: The increased risk of infection of disadvantaged individuals may also be due to the persistence of community transmission. These findings further highlight the need for interventions mitigating inequalities in the risk of SARS-CoV-2 infection and thus, of serious illness and mortality.

SeMPI 2.0-A Web Server for PKS and NRPS Predictions Combined with Metabolite Screening in Natural Product Databases.

Microorganisms produce secondary metabolites with a remarkable range of bioactive properties. The constantly increasing amount of published genomic data provides the opportunity for efficient identification of biosynthetic gene clusters by genome mining. On the other hand, for many natural products with resolved structures, the encoding biosynthetic gene clusters have not been identified yet. Of those secondary metabolites, the scaffolds of nonribosomal peptides and polyketides (type I modular) can be predicted due to their building block-like assembly. SeMPI v2 provides a comprehensive prediction pipeline, which includes the screening of the scaffold in publicly available natural compound databases. The screening algorithm was designed to detect homologous structures even for partial, incomplete clusters. The pipeline allows linking of gene clusters to known natural products and therefore also provides a metric to estimate the novelty of the cluster if a matching scaffold cannot be found. Whereas currently available tools attempt to provide comprehensive information about a wide range of gene clusters, SeMPI v2 aims to focus on precise predictions. Therefore, the cluster detection algorithm, including building block generation and domain substrate prediction, was thoroughly refined and benchmarked, to provide high-quality scaffold predictions. In a benchmark based on 559 gene clusters, SeMPI v2 achieved comparable or better results than antiSMASH v5. Additionally, the SeMPI v2 web server provides features that can help to further investigate a submitted gene cluster, such as the incorporation of a genome browser, and the possibility to modify a predicted scaffold in a workbench before the database screening.

Atlas of the immune cell repertoire in human atherosclerotic plaques characterized by single cell RNA-sequencing and multi-color flow cytometry

Abstract Rationale Atherosclerosis is a chronic inflammatory disease that is driven by the accumulation of pro- and anti-inflammatory leukocytes in the intima of affected arteries. Yet, the cellular composition of human atherosclerotic plaques is only poorly understood. Here, we characterized immune cells to human carotid atherosclerotic plaques by multi-color flow cytometry and scRNAseq. Methods and results First, we compared a set of previously reported digestion protocols to liberate leukocytes from human carotid plaques after surgical thrombendarteriectomy. One digestion cocktail, containing Collagenase IV and DNase I, was superior regarding cell survival and cell surface marker preservation. Second, leukocytes from 56 surgical specimen were characterized by flow cytometry with a set of 16 parameters and cell surface markers capable of identifying principal hematopoietic leukocyte lineages. This protocol allowed to extract and analyze on average 4x103 viable CD45+ leukocytes from a mean of 988 mg plaque tissue. Surprisingly, we found that atherosclerotic plaques were dominated by T cells with 33.7±2.2% CD4+ T-helper cells and 25.6±2.5% CD8+ cytotoxic T cells. CD11b+ myeloid cells, including monocytes and macrophages, represented only 20.2±4.0% of all CD45+ leukocytes. CD19+B cells and CD56+ NK-cells accounted for 3.9±1.2 and 3.3±0.5%, respectively. TCR-g/d+ T cells and neutrophils were undetectable in atherosclerotic plaques. This cellular composition differed significantly from peripheral blood, but was not relevantly changed between different plaque locations, indicating that macrophage-rich necrotic cores mostly contain dead cells. We confirmed the principal composition of human plaques by single-cell RNA-sequencing from six patients. To allow an estimation of cellular heterogeneity independent of classical cell surface marker assignment, we performed an unsupervised cluster detection algorithm by t-distributed stochastic neighbor embedding (tSNE) and found more than 16 leukocyte clusters with unique cell surface marker expression, suggesting an unexpected high diversity of plaque leukocytes. Conclusion We developed an immune cell phenotyping protocol optimized for human carotid plaques. The definition of phenotypes and frequencies in atherosclerotic plaques will allow to build clinical associations between the immune cell composition and clinical outcomes in future. Funding Acknowledgement Type of funding source: None

Unsupervised ranking of clustering algorithms by INFOMAX.

Clustering and community detection provide a concise way of extracting meaningful information from large datasets. An ever growing plethora of data clustering and community detection algorithms have been proposed. In this paper, we address the question of ranking the performance of clustering algorithms for a given dataset. We show that, for hard clustering and community detection, Linsker’s Infomax principle can be used to rank clustering algorithms. In brief, the algorithm that yields the highest value of the entropy of the partition, for a given number of clusters, is the best one. We show indeed, on a wide range of datasets of various sizes and topological structures, that the ranking provided by the entropy of the partition over a variety of partitioning algorithms is strongly correlated with the overlap with a ground truth partition The codes related to the project are available in https://github.com/Sandipan99/Ranking_cluster_algorithms.

Cluster Detection Mechanisms for Syndromic Surveillance Systems: Systematic Review and Framework Development.

BackgroundThe time lag in detecting disease outbreaks remains a threat to global health security. The advancement of technology has made health-related data and other indicator activities easily accessible for syndromic surveillance of various datasets. At the heart of disease surveillance lies the clustering algorithm, which groups data with similar characteristics (spatial, temporal, or both) to uncover significant disease outbreak. Despite these developments, there is a lack of updated reviews of trends and modelling options in cluster detection algorithms.ObjectiveOur purpose was to systematically review practically implemented disease surveillance clustering algorithms relating to temporal, spatial, and spatiotemporal clustering mechanisms for their usage and performance efficacies, and to develop an efficient cluster detection mechanism framework.MethodsWe conducted a systematic review exploring Google Scholar, ScienceDirect, PubMed, IEEE Xplore, ACM Digital Library, and Scopus. Between January and March 2018, we conducted the literature search for articles published to date in English in peer-reviewed journals. The main eligibility criteria were studies that (1) examined a practically implemented syndromic surveillance system with cluster detection mechanisms, including over-the-counter medication, school and work absenteeism, and disease surveillance relating to the presymptomatic stage; and (2) focused on surveillance of infectious diseases. We identified relevant articles using the title, keywords, and abstracts as a preliminary filter with the inclusion criteria, and then conducted a full-text review of the relevant articles. We then developed a framework for cluster detection mechanisms for various syndromic surveillance systems based on the review.ResultsThe search identified a total of 5936 articles. Removal of duplicates resulted in 5839 articles. After an initial review of the titles, we excluded 4165 articles, with 1674 remaining. Reading of abstracts and keywords eliminated 1549 further records. An in-depth assessment of the remaining 125 articles resulted in a total of 27 articles for inclusion in the review. The result indicated that various clustering and aberration detection algorithms have been empirically implemented or assessed with real data and tested. Based on the findings of the review, we subsequently developed a framework to include data processing, clustering and aberration detection, visualization, and alerts and alarms.ConclusionsThe review identified various algorithms that have been practically implemented and tested. These results might foster the development of effective and efficient cluster detection mechanisms in empirical syndromic surveillance systems relating to a broad spectrum of space, time, or space-time.

Multifaceted Traumatic Exposure: Simultaneous Direct and Vicarious Trauma Among EMS Personnel

Graph clustering is an important technique in data mining and network analysis, and it is widely used in chemistry, physics, biology, communication, and computer science. Similarities between vertices of a graph are the fundamental conditions for many hierarchical clustering algorithms. In the paper, we propose a new similarity measure based on microblock density, which computes the similarity between a pair of vertices by calculating the densities of their common adjacent microblock. This measure extends the scope and improves the discrimination of traditional measure, thus significantly improving the performance and stability of the similarity-based clustering algorithms. Experiments on synthetic data and real networks show that the density-based similarity approach accurately reflects the local structure of the graph and provides higher accuracy similarities for clustering and community structural detection algorithms than other state-of-the-art methods.

On the measurements of assembly bias and splashback radius using optically selected galaxy clusters

ABSTRACT We critically examine the methodology behind the claimed observational detection of halo assembly bias using optically selected galaxy clusters by Miyatake et al. and More et al. We mimic the optical cluster detection algorithm and apply it to two different mock catalogues generated from the Millennium simulation galaxy catalogue, one in which halo assembly bias signal is present, while the other in which the assembly bias signal has been expressly erased. We split each of these cluster samples into two using the average cluster-centric distance of the member galaxies to measure the difference in the clustering strength of the subsamples with respect to each other. We observe that the subsamples split by cluster-centric radii show differences in clustering strength, even in the catalogue where the true assembly bias signal was erased. We show that this is a result of contamination of the member galaxy sample from interlopers along the line of sight. This undoubtedly shows that the particular methodology adopted in the previous studies cannot be used to claim a detection of the assembly bias signal. We figure out the tell-tale signatures of such contamination, and show that the observational data also show similar signatures. Furthermore, we also show that projection effects in optical galaxy clusters can bias the inference of the 3D edges of galaxy clusters (splashback radius), so appropriate care should be taken while interpreting the splashback radius of optical clusters.

Association of Microcalcification Clusters with Short-term Invasive Breast Cancer Risk and Breast Cancer Risk Factors

Using for-presentation and for-processing digital mammograms, the presence of microcalcifications has been shown to be associated with short-term risk of breast cancer. In a previous article we developed an algorithm for microcalcification cluster detection from for-presentation digital mammograms. Here, we focus on digitised mammograms and use a three-step algorithm. In total, 253 incident invasive breast cancer cases (with a negative mammogram between three months and two years before diagnosis, from which we measured microcalcifications) and 728 controls (also with prior mammograms) were included in a short-term risk study. After adjusting for potential confounding variables, we found evidence of an association between the number of microcalcification clusters and short-term (within 3–24 months) invasive breast cancer risk (per cluster OR = 1.30, 95% CI = (1.11, 1.53)). Using the 728 postmenopausal healthy controls, we also examined association of microcalcification clusters with reproductive factors and other established breast cancer risk factors. Age was positively associated with the presence of microcalcification clusters (p = 4 × 10−04). Of ten other risk factors that we studied, life time breastfeeding duration had the strongest evidence of association with the presence of microcalcifications (positively associated, unadjusted p = 0.001). Developing algorithms, such as ours, which can be applied on both digitised and digital mammograms (in particular for presentation images), is important because large epidemiological studies, for deriving markers of (clinical) risk prediction of breast cancer and prognosis, can be based on images from these different formats.

Simulation of solute clusters in metallic systems

The arrangement of atoms in the microstructures of alloys plays a key role in defining their physical and functional properties. These microstructures are studied to detect deviations from the expected homogeneous distribution using frequency distribution tests and cluster detection algorithms on reconstructions obtained from atom probe microscopy. However, it is challenging to evaluate and validate the performance of these algorithms using the reconstructed data because of the experimental limitations of the atom probe data itself. Low detector efficiency and reconstructional artifacts may lead to loss of existing clusters, or introduce heterogeneities that do not reflect the original arrangement. Additionally, retrieving atom probe data is a time and cost-prohibitive process. These challenges call for the use of simulated datasets with known concentrations and distributions to evaluate algorithms that analyze distributions or detect clusters. In this paper, we extend the existing methodology to simulate the creation of solute clusters in very dilute solutions for any solution. We propose methods to create denser and wider clusters in any solution irrespective of the concentration. Three novel solute cluster simulation techniques incremental, hierarchical and gravitational clustering are presented and compared in this paper. The simulated datasets can be used to validate and compare the performance of frequency distribution or cluster discovery algorithms. We compare the cluster analysis results obtained from the single linkage algorithm and DBSCAN to highlight the utility of these datasets.

SoyCSN: Soybean context-specific network analysis and prediction based on tissue-specific transcriptome data.

The Soybean Gene Atlas project provides a comprehensive map for understanding gene expression patterns in major soybean tissues from flower, root, leaf, nodule, seed, and shoot and stem. The RNA‐Seq data generated in the project serve as a valuable resource for discovering tissue‐specific transcriptome behavior of soybean genes in different tissues. We developed a computational pipeline for Soybean context‐specific network (SoyCSN) inference with a suite of prediction tools to analyze, annotate, retrieve, and visualize soybean context‐specific networks at both transcriptome and interactome levels. BicMix and Cross‐Conditions Cluster Detection algorithms were applied to detect modules based on co‐expression relationships across all the tissues. Soybean context‐specific interactomes were predicted by combining soybean tissue gene expression and protein–protein interaction data. Functional analyses of these predicted networks provide insights into soybean tissue specificities. For example, under symbiotic, nitrogen‐fixing conditions, the constructed soybean leaf network highlights the connection between the photosynthesis function and rhizobium–legume symbiosis. SoyCSN data and all its results are publicly available via an interactive web service within the Soybean Knowledge Base (SoyKB) at http://soykb.org/SoyCSN. SoyCSN provides a useful web‐based access for exploring context specificities systematically in gene regulatory mechanisms and gene relationships for soybean researchers and molecular breeders.

Data Mining Based Outlier Cluster Detection Algorithm

Outlier detection has been engaged and considered field as of late. Exception recognition is a significant information mining movement with various applications, including MasterCard misrepresentation location, disclosure of crimes in electronic trade, video reconnaissance, climate forecast, and pharmaceutical re-search. The assignment of anomaly recognition is to distinguish the information objects from heaps of all objects. Outlier mining normally utilized in different fields. The vast majority of the strategies gives more regard for distinguish information objects from worldwide view which is in suitable for multidimensional information sets. According to technology, outlier recognition will be troublesome by utilizing conventional arrangements. In this, we are utilizing two calculations, for example, ODA and LOF. ODA used to locate the mean or exceptions from multi-dimensional informational indexes. LOF used to discover the likelihood of that anomalies utilizing thickness based calculations. Exception location is a significant undertaking in information mining. Nearby anomalies contrasting with their nearby neighborhood rather than worldwide information. Anomaly discovery is an essential assignment in information mining. Nearby exceptions contrasting with their neighborhood close articles rather than worldwide information dissemination. The procedure segmented into two phases, online and disconnected. Small scale groups are made in online stage and last bunches are produced in disconnected stage. Exception location is a significant assignment in information mining. Nearby exceptions contrasting with their neighborhood rather than worldwide information. Relative thickness of an item against its neighbors as the marker of the level of the article being anomaly which is allotted to nearby exception factor.

Euclid preparation

Galaxy cluster counts in bins of mass and redshift have been shown to be a competitive probe to test cosmological models. This method requires an efficient blind detection of clusters from surveys with a well-known selection function and robust mass estimates, which is particularly challenging at high redshift. The Euclid wide survey will cover 15 000 deg2 of the sky, avoiding contamination by light from our Galaxy and our solar system in the optical and near-infrared bands, down to magnitude 24 in the H-band. The resulting data will make it possible to detect a large number of galaxy clusters spanning a wide-range of masses up to redshift ∼2 and possibly higher. This paper presents the final results of the Euclid Cluster Finder Challenge (CFC), fourth in a series of similar challenges. The objective of these challenges was to select the cluster detection algorithms that best meet the requirements of the Euclid mission. The final CFC included six independent detection algorithms, based on different techniques, such as photometric redshift tomography, optimal filtering, hierarchical approach, wavelet and friend-of-friends algorithms. These algorithms were blindly applied to a mock galaxy catalog with representative Euclid-like properties. The relative performance of the algorithms was assessed by matching the resulting detections to known clusters in the simulations down to masses of M200 ∼ 1013.25 M⊙. Several matching procedures were tested, thus making it possible to estimate the associated systematic effects on completeness to < 3%. All the tested algorithms are very competitive in terms of performance, with three of them reaching > 80% completeness for a mean purity of 80% down to masses of 1014 M⊙ and up to redshift z = 2. Based on these results, two algorithms were selected to be implemented in the Euclid pipeline, the Adaptive Matched Identifier of Clustered Objects (AMICO) code, based on matched filtering, and the PZWav code, based on an adaptive wavelet approach.

Spatial cluster detection of regression coefficients in a mixed‐effects model

AbstractIdentifying spatial clusters of different regression coefficients is a useful tool for discerning the distinctive relationship between a response and covariates in space. Most of the existing cluster detection methods aim to identify the spatial similarity in responses, and the standard cluster detection algorithm assumes independent spatial units. However, the response variables are spatially correlated in many environmental applications. We propose a mixed‐effects model for spatial cluster detection that takes spatial correlation into account. Compared to a fixed‐effects model, the introduced random effects explain extra variability among the spatial responses beyond the cluster effect, thus reducing the false positive rate. The developed method exploits a sequential searching scheme and is able to identify multiple potentially overlapping clusters. We use simulation studies to evaluate the performance of our proposed method in terms of the true and false positive rates of a known cluster and the identification of multiple known clusters. We apply our proposed methodology to particulate matter (PM2.5) concentration data from the Northeastern United States in order to study the weather effect on PM2.5 and to investigate the association between the simulations from a numerical model and the satellite‐derived aerosol optical depth data. We find geographical hot spots that show distinct features, comparing to the background.

A Comparative Study of Cluster Detection Algorithms in Protein-Protein Interaction for Drug Target Discovery and Drug Repurposing.

The interactions between drugs and their target proteins induce altered expression of genes involved in complex intracellular networks. The properties of these functional network modules are critical for the identification of drug targets, for drug repurposing, and for understanding the underlying mode of action of the drug. The topological modules generated by a computational approach are defined as functional clusters. However, the functions inferred for these topological modules extracted from a large-scale molecular interaction network, such as a protein–protein interaction (PPI) network, could differ depending on different cluster detection algorithms. Moreover, the dynamic gene expression profiles among tissues or cell types causes differential functional interaction patterns between the molecular components. Thus, the connections in the PPI network should be modified by the transcriptomic landscape of specific cell lines before producing topological clusters. Here, we systematically investigated the clusters of a cell-based PPI network by using four cluster detection algorithms. We subsequently compared the performance of these algorithms for target gene prediction, which integrates gene perturbation data with the cell-based PPI network using two drug target prioritization methods, shortest path and diffusion correlation. In addition, we validated the proportion of perturbed genes in clusters by finding candidate anti-breast cancer drugs and confirming our predictions using literature evidence and cases in the ClinicalTrials.gov. Our results indicate that the Walktrap (CW) clustering algorithm achieved the best performance overall in our comparative study.

Data Mining Based Outlier Cluster Detection Algorithm

Outlier detection has been engaged and considered field as of late. Exception recognition is a significant information mining movement with various applications, including MasterCard misrepresentation location, disclosure of crimes in electronic trade, video reconnaissance, climate forecast, and pharmaceutical re-search. The assignment of anomaly recognition is to distinguish the information objects from heaps of all objects. Outlier mining normally utilized in different fields. The vast majority of the strategies gives more regard for distinguish information objects from worldwide view which is in suitable for multidimensional information sets. According to technology, outlier recognition will be troublesome by utilizing conventional arrangements. In this, we are utilizing two calculations, for example, ODA and LOF. ODA used to locate the mean or exceptions from multi-dimensional informational indexes. LOF used to discover the likelihood of that anomalies utilizing thickness based calculations. Exception location is a significant undertaking in information mining. Nearby anomalies contrasting with their nearby neighborhood rather than worldwide information. Anomaly discovery is an essential assignment in information mining. Nearby exceptions contrasting with their neighborhood close articles rather than worldwide information dissemination. The procedure segmented into two phases, online and disconnected. Small scale groups are made in online stage and last bunches are produced in disconnected stage. Exception location is a significant assignment in information mining. Nearby exceptions contrasting with their neighborhood rather than worldwide information. Relative thickness of an item against its neighbors as the marker of the level of the article being anomaly which is allotted to nearby exception factor.

Multiple failure behaviors identification and remaining useful life prediction of ball bearings

Accurate remaining useful life (RUL) prediction is the key for successful implementation of condition based maintenance program in any industry. Data driven prognostics approaches are generally used to predict the RUL of the components. Presence of noise in the data reduces the accuracy of RUL prediction. Mechanical components are prone to failures due to several failure modes; resulting into multiple failure behaviors or patterns in life test data obtained from various units. If such failure patterns or behaviors are not identified and treated appropriately, the same may act as one of the sources for data noise. In the present research, clustering and change point detection algorithm (CPDA) is used for identification of the presence of multiple failure behaviors in the data. Silhouette width value is used to find out the optimum number of clusters. Combined output of clustering and CPDA is used for developing RUL prediction models. Separate models for single and multiple failure behaviors are constructed using General Log-Linear Weibull (GLL-Weibull) distribution. Results show that identification of failure behavior helps in accurate prediction of RUL. The approach is validated using roller ball bearing life test data.

Meteor shower detection with density‐based clustering

AbstractWe present a new method to detect meteor showers using the density‐based spatial clustering of applications with noise algorithm (DBSCAN; Ester et al. ). The DBSCAN algorithm is a modern cluster detection algorithm that is well suited to the problem of extracting meteor showers from all‐sky camera data because of its ability to efficiently extract clusters of different shapes and sizes from large data sets. We apply this shower detection algorithm on a data set that contains 25,885 meteor trajectories and orbits obtained from the NASA All‐Sky Fireball Network and the Southern Ontario Meteor Network (SOMN). Using a distance metric based on solar longitude, geocentric velocity, and Sun‐centered ecliptic radiant, we find 25 strong cluster detections and six weak detections in the data, all of which are good matches to known showers. We include measurement errors in our analysis to quantify the reliability of cluster occurrence and the probability that each meteor belongs to a given cluster. We validate our method through false‐positive/negative analysis and with a comparison to an established shower detection algorithm.

A novel outlier cluster detection algorithm without top-n parameter

Outlier detection is an important task in data mining with numerous applications, including credit card fraud detection, video surveillance, etc. Outlier detection has been widely focused and studied in recent years. The concept about outlier factor of object is extended to the case of cluster. Although many outlier detection algorithms have been proposed, most of them face the top-n problem, i.e., it is difficult to know how many points in a database are outliers. In this paper we propose a novel outlier cluster detection algorithm called ROCF based on the concept of mutual neighbor graph and on the idea that the size of outlier clusters is usually much smaller than the normal clusters. ROCF can automatically figure out the outlier rate of a database and effectively detect the outliers and outlier clusters without top-n parameter. The formal analysis and experiments show that this method can achieve good performance in outlier detection.