+91-9519066910

Assignment

Computing Skills Assignment Help

Matlab Assignment Help

An IncrementalCFSAlgorithm

Rating:
An IncrementalCFSAlgorithm

An IncrementalCFSAlgorithm forClustering LargeData in

Industrial InternetofThings QingchenZhang,ChunshengZhu,LaurenceT.Yang,ZhikuiChen,LiangZhao, andPeng Li Abstract—

With the rapid advances of sensing technolo- enables conditionmonitoring, structural healthmonitoring, re- gies and wireless communications, large amounts of dy- motediagnosis,and remotecontrolofproductionsystems in real namic data pertaining to industrial production are being time.Furthermore, industrial IoTmakes smart factoriesbecomecollected frommany sensor nodes deployed in the indus- trial InternetofThings.Analyzing thosedataeffectivelycan possible fordynamicallyorganizing andoptimizingproduction help to improve the industrialservicesandmitigate thesys- processes [2]. tem unprepared breakdowns.

As an important technique Meanwhile, industrial IoT bringsmany new challengeswith of data analysis, clustering attempts to find the underlying regard to smart control and large datamanagement [3]. Espe- pattern structures embedded inunlabeled information.Un- cially, a lot of sensors deployed in industrial IoT systems arefortunately,most of the current clustering techniques that could only dealwith static data become infeasible to clus-

collecting a significant volume of data stream or dynamic data ter a significant volume of data in the dynamic industrial [4]. However, it is a challenging issue to analyze large data applications. To tackle this problem, an incremental clus- stream since large numbers of samples arriving in a stream tering algorithm by fast finding and searching of density areusually timedependent, and theunderlyingpattern that they peaksbasedonk-mediods isproposed in thispaper. In the presentmayevolveover time.Analyzing thosecollecteddataef-proposedalgorithm, twoclusteroperations,namelycluster creating and clustermerging, are defined to integrate the fectivelycanhelp to improve the industrialservicesandmitigate currentpattern into thepreviousone for thefinalclustering the systemunprepared risks.Forexample, ifuserscancorrectly result, and k-mediods is employed tomodify the clustering analyze thecollected informationof thephysicaldevicessuchas centersaccording to thenewarrivingobjects

.Finally,exper- temperatures, pressures, and gas compositions by thousands of iments are conducted to validate the proposed scheme on three popularUCI datasets and two real datasets collected sensorsdeployed ina largegas turbine, theycan take immediate from industrial InternetofThings in termsof clustering ac- actions to prevent the output of electricity before the systems curacyandcomputational time. breaks downwhen it happens to be abnormal.Recently, some methods and platforms for big data stream analysis have been Index Terms—CFS clustering, incremental clustering, in- dustrial InternetofThings (IoT),K-mediods. developed formany scientific and industrial applications. For example, Cao et al. [5] proposed a general framework to dis- I. INTRODUCTION cover distance-based outliers from huge volumes of streaming VER the last few years,with the rapid advances of sens- databy combining twogeneraloptimizationprinciples,namely ing technologiesandwirelesscommunications, industrial theminimalprobingprincipleand the lifespan-awareprioritiza- IOnternet of Things (IoT) hasmade a great progress

[1]. Indus- tion principle.Aiming at the classification problem of big data trial IoT is created by embedding smart electronics into pro- stream,DeRosa et al. [6] developed an incremental and non- duction systems via a dynamic global information network. It parametric approach, whichworks by incrementally covering is improving the effectiveness and efficiency ofmodern indus- the input space and dynamically accommodating new classes trial production and applications. For example, industrial IoT when they appear in thedata stream.Fong etal. [7]presented a novel lightweight featureselectionalgorithm forbigdatastream miningbyusingacceleratedparticle swarmoptimization (PSO)Manuscript receivedNovember8,2015; revisedMarch23,2016,June 28,2016,January29,2017,andFebruary17,2017;acceptedMarch14, type of swarm. In industrial applications such asGoogle and 2017.Date of publicationMarch 20, 2017; date of current version June Amazon, cloud computing platforms, like Hadoop and Ama- 1,2017.Paperno.TII-15-1648.R4. (Corresponding author: Laurence T. zonsEC2, are used to analyze big data stream, by providing aYang.)

Q. Zhang and L. T.

Yang arewith theSchool ofElectronicEngineer- largenumberof computation and storage resources.Especially, ing,UniversityofElectronicScienceandTechnologyofChina,Chengdu anefficientcloudcomputingprogrammingmodel,MapReduce, 611731, China, and also with the Department of Computer Science, wasproposed forbigdatacomputation.Using thismodel,usersSt.FrancisXavierUniversity,AntigonishNSB2G2W5,Canada (e-mail: qingchen@mail.dlut.edu.cn; ltyang@gmail.com). only need to define theMap function and theReduce function C.Zhu iswith theDepartmentofElectricalandComputerEngineering, to implement distributed algorithms efficiently [8]. This paper UniversityofBritishColumbia,Vancouver,BCV6T1Z4,Canada (e-mail: focuseson theclustering technology fordata stream indynamicchunsheng.tom.zhu@gmail.com). Z. Chen, L. Zhao, and P. Li are with the School of Software industrial IoT. Technology, Dalian University of Technology, Dalian 116024, China As one important technique of data analysis, clustering aims (e-mail: zkchen@dlut.edu.cn; matthew1988zhao@mail.dlut.edu.cn; topartition adataset into several clusters according to themea-lipeng2015@mail.dlut.edu.cn). DigitalObject Identifier 10.1109/TII.2017.2684807 sured similarity such that the objects in the same cluster are 1551-3203© 2017 IEEE.Personal use is permitted, but republication/redistribution requires IEEEpermission. See http://www.ieee.org/publications standards/publications/rights/index.html formore information.

themostwell-known incremental scheme based on biological intelligence is the incremental ant colony clustering [17].

This algorithmuses artificial ants topickupordropdownobjects to implementclusteringaccording to thesimilaritywithother local regional objects. Then, amechanism of decomposing clusters wasused to formnewclusterswhenusers interestsarechanged. Li [18] proposed an incremental clustering algorithm based on chaos and immune response. In this algorithm, the diversity of the chaotic sequence is used to recognize the incremental data which do not belong to any existing clusters.Moreover, the memory antibodies produced by the primary immune response are used for the secondary immune response to recognize the incremental data which belong to the existing cluster. Other methods of this type include incremental clustering based on thePSO and the artificial immune system [19], [20].However, they are of high computational complexity due to the introduc- tion of intelligent techniques, so they are still not suitable for industrialdynamicdata.To tackle thisproblem, two clusterop- erations, i.e.,clustercreatingandclustermerging,aredefined to improve clustering efficiency for integrating the currentpattern algorithm for big data clustering. However, CFS clustering into theprevious one in thispaper. is initially designed for static data. This paper focuses on the extension of the algorithm to handle dynamic data,making it efficient enough tobe used in industrial applications. III.

INCREMENTALCFSCLUSTERINGBASEDON K-MEDIODS

A. Problem Formulation

B. Incremental Clustering Given a sequentially collecteddataset

Xt ,t = 1,2,... ,T, In recent years, with the increasing popularity of dynamic whereXt consistsofnt objects, eachwit{hm}featuresobserved data, many incremental clustering methods have been devel- at the timestamp t, thegoalof incrementalclustering is tomain- oped, which can be roughly categorized by two groups: the tain a set of k clusters such that when some new objects are variations of traditional clustering techniques and incremental coming, either assigning them into the current k clusters, or schemes based on biological intelligence. creating one ormore new clusters. Incremental clustering al- A representative example of the first type is incremental gorithms are divided into two types: single-pass incremental density-based clustering,which focuseson analysisofdynamic clustering algorithms and cluster center adding algorithms.The data in the data warehouse [14]. Especially, the performance former assigns the new objects into the current clusters and of this method was estimated by a series of experiments on then redefines the cluster centers accordingly, so thenumber of a large of datasets. This kind of incremental schemes need to clusters does not change. In the latter, new clusters are created storealldataobjects in thememory forpreprocessing, resulting continuallyand thenumberofclustersusually increasesasmore in the limited performance for clustering large data.

Another andmoreobjects arrive. example is incremental fuzzy c-means algorithm. Incremental In this paper, the second type is studied and an incremen- fuzzy c-means clustering processes data chunk by chunk and tal CFS algorithm is proposed for industrial dynamic data estimates the c-centroids for the entire dataset by extracting clustering. Assume that the set of all available objects is the information in each chunk.Theirkernel versionshavebeen Ut 1=X1 X2 Xt 1 at the timestamp t

1 and the − −also developed to cluster large data efficiently [10].Moreover,

correspondin∪gclus∪te·r·in·g∪result isRt 1.At the times−tampt,CFS −two algorithms based on fuzzy c-medoids called online fuzzy isused tocluster thedatasetXt for the resultct.The incremental c-medoids andhistory-basedonline fuzzy c-medoids aredevel- CFS clustering aims to integrate the result ct into the previous oped to cluster large relational datasets [11]. However, these oneRt 1 for thefinalpatternRt with regard to all the available −algorithms require predefining the number of clusters,making objectsUt =Ut 1 Xt. −them infeasible for industrial applications.To tackle this issue, According to the∪above analysis, there are twomajor chal- an incremental CFS clustering algorithm based on k-mediods lenges for implementing the incrementalCFS clustering algo- is proposed, in which CFS is used to determine the number rithm. The first challenge is how to integrate the result ct into of clusters automatically and k-mediods is used tomodify the the previous oneRt 1 for the final patternRt. The other one −clustering centers according to thenew arriving objects. is how to update the clustering centers depending on the new Thebiological intelligence-based clustering enhances the ro- arriving objects. In particular, each object in the previous set bustnessof traditionalclusteringalgorithmsbycombining them Ut 1 has its value of γ,while each new arriving object is still −with intelligent techniques.A largenumberofexperimentshave at the initial levelwith the value of zero. All available objects shown their effectiveness in avoiding localminimum.Perhaps, cannot be loaded into thememory, so it is impossible to find

marketing assignmentmarketing management assignmentmarketing management assignment helpdigital marketing assignment helpmarketing assignment help onlinemarketing homework helpmarketing assignment writingonline marketing assignment helpstrategic marketing assignment helponline marketing assignmentmarketing assignment writing serviceinternational marketing assignment helpmarketing mix assignment helpmarketing homework assignment

 



 



 DOWNLOAD SAMPLE ANSWER


   

Details

  • Number of views:
    146
  • PRICE :
    AU$ 0.00

Submit Works

Drop File To Upload Or
BROWSE
Assignment Help Australia

Our Top Experts


Karen Betty

Holding a PhD degree in Finance, Dr. John Adams is experienced in assisting students who are in dire need...

55 - Completed Orders


Daphne Lip

Canada, Toronto I have acquired my degree from Campion College at the University of Regina Occuption/Desi...

52 - Completed Orders


Mr Roberto Tuzii

Even since I was a student in Italy I had a passion for languages, in fact I love teaching Italian, and I...

102 - Completed Orders


Harsh Gupta

To work with an organization where I can optimally utilize my knowledge and skills for meeting challenges...

109 - Completed Orders


ARNAB BANERJEE

JOB OBJECTIVE Seeking entry level assignments in Marketing & Business Development with an organization...

202 - Completed Orders


KARAN BHANDARI

Current work profile Project manager- The Researchers Hub (2nd Jan 2016 to presently working) Researc...

20 - Completed Orders


Tan Kumar Ali

Sales Assistant, Mito Marina Assigned to the Stationery dept – assisted in merchandising, stock taking...

100 - Completed Orders


Wesenu Irko

Personal Profile Dedicated and highly experienced private chauffeur. High energy, hardworking, punctua...

200 - Completed Orders