research article a new classification analysis of customer...

Research ArticleA New Classification Analysis of CustomerRequirement Information Based on QuantitativeStandardization for Product Configuration

Zheng Xiao Zude Zhou and Buyun Sheng

School of Mechanical and Electronic Engineering Wuhan University of Technology No 122 Luoshi RoadHongshan Wuhan Hubei 430070 China

Correspondence should be addressed to Zheng Xiao reallylaughwhuteducn

Received 13 October 2015 Accepted 30 December 2015

Academic Editor David Bigaud

Copyright copy 2016 Zheng Xiao et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

Traditional methods used for the classification of customer requirement information are typically based on specific indicatorshierarchical structures and data formats and involve a qualitative analysis in terms of stationary patterns Because these methodsneither consider the scalability of classification results nor do they regard subsequent application to product configuration theirclassification becomes an isolated operation However the transformation of customer requirement information into quantifiablevalues would lead to a dynamic classification according to specific conditions and would enable an association with productconfiguration in an enterprise This paper introduces a classification analysis based on quantitative standardization which focuseson (i) expressing customer requirement information mathematically and (ii) classifying customer requirement information forproduct configuration purposes Our classification analysis treated customer requirement information as follows first it wastransformed into standardized values usingmathematics subsequent to which it was classified through calculating the dissimilaritywith general customer requirement information related to the product family Finally a case study was used to demonstrate andvalidate the feasibility and effectiveness of the classification analysis

1 Introduction

In the past decades the classification of customer require-ment information (CRI) has become increasingly importantin the entire product development and manufacturing pro-cess Numerous researchers have accomplished valuableworkin terms of developing approaches to classification and threeof these methods that is those based on the Kano modelproduct hierarchical structure and data format are the mostprominent

The Kano model was introduced by Noriaki Kano whoclassified customer preferences according to a threshold per-formance and excitement to guide design decisions [1] TheKano model is often associated with quality function deploy-ment (QFD) and fuzzy mathematics which was utilized todetermine the weights or importance of customer require-ments [2 3]The product hierarchical structure classifies CRIinto different types according tomodern product hierarchies

such as function form extension and price requirements[4 5] and is a form of qualitative analysis intended to provideperformance indicators The data format divides CRI intobinary option parameter description and interpretationdata types in terms of common data formats on the Web[6 7] Compared with the product hierarchical structure andthe KANOmodel the classification based on the data formathas already solved the problem of quantitative expression andanalysis for CRI analysis Moreover the premise of a classifi-cation based on the data format is to collect CRI via theWeban approach that is already adopted and is likely to be widelyutilized in the future [8] The classification of CRI using thedata format facilitates not only the subsequent analysis of theinformation but also the management thereof Neverthelessclassification based on the data format has certain limitationsbecause this method does not take into account the numberof data formats that exist on theWeb or how this classificationrelates to the entire product configuration process

Hindawi Publishing CorporationMathematical Problems in EngineeringVolume 2016 Article ID 7274538 8 pageshttpdxdoiorg10115520167274538

2 Mathematical Problems in Engineering

In conclusion traditional classificationmethods focus oncustomer requirements rather than on product configurationsuch that the classification results are neither extendable norscalable The processes of CRI classification and productconfiguration can occur independently of one another Forinstance customersrsquo preferences are able to complete trans-formation into product design [9 10] without performingCRI classification although CRI classification results areuseful for determining them Additionally the lack of mathe-matical analysis of CRImeans that it is difficult to use the clas-sification results directly for product configuration purposesTherefore the modern approach to CRI classification has notbeen a simple issue of dividing the information into groupsinstead it has to consider (i) how to analyze and express CRIwith mathematical methods and (ii) how to classify CRI forthe purpose of product configuration

In this paper we introduce a classification analysis basedon quantitative standardization Section 2 describes a math-ematical model to analyze CRI in terms of product fami-lies Section 3 provides details as to how to transform CRIinto quantitative standards On the basis of Sections 2 and3 Section 4 presents the classification method Finally inSection 5 a case is demonstrated to confirm the feasibility andeffectiveness of the proposed method

2 Background Review

21 CRI Structured Procedure The increasing application ofe-commerce has been transforming the acquisition methodof choice from offline to online Online CRI acquisitionmainly depends on theWeb including the advantages it offersin terms of efficiency affordability and convenience [11]Online CRI can be submitted in the form of XMLdocumentswhich can containmultiple kinds of data and can adapt to thedynamic development of CRI AnXMLdocument containingCRI is referred to as a customer requirement informationdocument (CRID) Because of the XML framework eachCRID can tag various CRI features of each customer how-ever these features contain multiple data types character-ized by fuzziness concealment and similarity such thatit is difficult to correctly identify the information for use inthe process of product configuration Thus it is necessary totransform CRI into a structured model corresponding to theproduct family model to enable the information to be trans-lated into product development for manufacturing Researchon the structure model of the product family has led to theconstruction of general customer requirement information(GCRI) [12] which is able to abstract a series of similarCRI featureswhose personalized features are distinguished byspecific values of cases [13] Figure 1 illustrates the standard-ization procedure in which CRI is transformed into GCRIwhich means that features with values in CRID categoriescan be transformed into the corresponding GCRI classes in astructured model

22 CRI Document Model In this model the CRI is submit-ted in the form of CRID which is a document based on adocument representation model capable of enlightening the

CRID1CRID2

CRID3CRI structure model

C11 C12 C21 C22 C23 C31 C32

GCRI (F1) GCRI (F2) GCRI (F3)

middot middot middot

Figure 1 Summary of the standardization procedure

CRID representation Because a document is composed ofwords theword is themostwidely used unit of information indocumentmodeling [14] namely a document representationmodel can be established by using the characteristics of wordsand is implemented by the Vector Space Model (VSM) TheVSM [15] is a vector model utilizing the extracted character-istics of words from a document in a Euclidean space [16] Aword characteristic corresponds to a separate term If a termoccurs in the document its value in the vector is nonzero [17]The main idea of VSM is as follows let 119863 = [119863

1 1198632 119863

119889]

be a document set with each document 119863119894(119894 = 1 2 119889)

represented by a set of terms 119879 = [1198791 1198792 119879

119905] any

119879119895(119895 = 1 2 119905) corresponds to one dimension in the

VSM such that 119863 can be a 119889 times 119905 matrix which means thatdocuments can be mapped to a point in the VSM and theirsimilarity can be calculated by distances To date the VSM isthe most efficient and useful document representation modelbecause it transforms the similarity between two documentsinto the similarity between two vectors [18]

Thus the representation model of the CRID is a one-dimensional vector containing the values of the CRI featuresIf the number of CRID tags is 119899 and the CRI features areindependent of each other anyCRID119863

119894can be represented as

a one-dimensional set [1199031198941 1199031198942 119903

119894119899] On the basis of this if

the number of CRIDs is119898 and each CRID defines 119899 featuresthere will be an119898 times 119899matrix shown as follows

119877 =

[[[[[[

[

11990311

11990312

sdot sdot sdot 1199031119899

11990321

11990322


1199031198981

1199031198982

sdot sdot sdot 119903119898119899

]]]]]]

]

(1)

In the matrix 119877 119903119894119895(119894 = 1 2 119898 119895 = 1 2 119899)

is a 119895 feature value submitted by one customer 119894 Howeveran existing problem of CRI features is that the informationmay occur as a combination of multiple data types such asnumbers andwordsThis issue would have to be addressed byanalyzing the similarity of these values to introduce quantita-tive standardization of CRI such that 119903

119894119895in matrix 119877 could be

changed into a uniform fundamental unit [19]

23 Classification and Clustering Classification and clus-tering are two major methods for information analysisespecially data in XML documents [20] The aim of theclassification is to build a classifier based on some cases withsome attributes to describe the objects or one attribute todescribe the group of the objectsThen the classifier is used to

Mathematical Problems in Engineering 3

CRI feature variables Quantitative analysis Standard transform CRI standardized values

Figure 2 Process of CRI quantitative standardization

predict the group attributes of new cases based on the valuesof other attributes [21]The aim of clustering is to find groupsof objects such that the objects in a group will be similar toone another and different from the objects in other groupsThe clustering algorithm has access only to the set of featuresdescribing each object it is not given any labels as to whereeach of the instances should be placed within the partition[22] Thus classification is supervised learning as targets arepredefined whereas clustering is generally used in an unsu-pervised fashion

The typical clustering algorithm is 119870-means [23] whichaims to partition 119899 observations into 119896 clusters Each obser-vation belongs to the cluster with the nearest mean Since thesum of squares is the squared Euclidean distance [24] this isintuitively the ldquonearestrdquo mean 119870-means clustering is able tocompute fast and compatiblewithmassive dataHowever dueto its unsupervised fashion there must be an issue of how tochoose 119896 and centroids Subsequent research on semisuper-vised clustering [25] is to remedy this defect because guiding aclustering algorithm is very efficient for improving its quality

In practice if the centroids and 119896 can be defined bysome technique indicators a classifier will be built based ondissimilarity calculation among objects In the next sectionsaiming at CRI the construction of this classifier will beintroduced

3 Quantitative Standardization of CRI

The CRI features contain multitype data values in a CRIDincluding nominal and scaled variables The nominal vari-ables are binary and multiple and the scaled variables aremeasured The process of CRI quantitative standardization isshown in Figure 2 and has the purpose of realizing quantifi-cation for the assignment of uniform fundamental units ofCRI feature values

31 Quantitative Analysis Thenominal binary variables havethe values of 0 and 1 where 0 or 1 means a CRI feature valuedoes not exist or exists respectively The scaled variables arereal values with fundamental units The nominal multiplevariables are different from the former two which may notcorrespond to real values with multiple states Thus theremust be a quantitative analysis for variable assignment

The differences among the various features and thedifferences between the feature states are accounted for byproposing methods based on fuzzy mathematics for the pur-pose of linguistic representation for the quantitative analysis

(i) Quantitative Analysis Based on Fuzzy Mathematics Let 119880be a discussion universe if mapping 120583

119860

119880 rarr [0 1] 119906 997891rarr

120583119860

(119906) isin [0 1] defines a fuzzy set119860in119880 120583

119860

(119906)will be named

the membership function of the fuzzy set 119860[26] which can

be expressed as

119860= (120583119860

(1199061) 120583119860

(1199062) 120583

119860

(119906119899)) (2)

The defuzzification calculation formula of 119860is

119860 = radic120583119860

(1199061)2

+ 120583119860

(1199062)2

+ sdot sdot sdot + 120583119860

(119906119899)2

(3)

The membership function is also represented as a fuzzydistribution of which the trapezoidal distribution is com-monly used The trapezoidal distribution 119872

is described by

four parameters119872

= (119897 1198981 1198982 119899) whose fuzzymembership

function can be expressed as

120583119872

(119906) =

119906 minus 119897

1198981minus 119897

119897 le 119906 le 1198981

1 1198981le 119906 le 119898

2

119906 minus 119899

1198982minus 119899

1198982le 119906 le 119899

0 others

(4)

According to Chenrsquos model [27] the defuzzification calcula-tion formula of119872

is

119872 =119897 + 1198981+ 1198982+ 119899

4 (5)

(ii) Quantitative Analysis Based on Linguistic RepresentationThe linguistic representation is achieved by constructinga fuzzy function with a linguistic evaluation scale rule inTable 1 In terms of the number of nominal multiple variablestates the number of linguistic terms is determined so that alinguistic set NVL LRL LLM LHRHHVHP can bedefined Finally the linguistic set is transformed into fuzzyfunctions by linguistic representation [28 29]

32 Standard Transform The similarity of quantitative CRIwith different fundamental units is measured by using astandard transform formula

119903119894119895=

119903119894119895minus 119903119895

radic(1 (119898 minus 1))sum119898

119894=1(119903119894119895minus 119903119895)2

isin [minus1 1]

119903lowast

119894119895=

119903119894119895minusmin 119903

119894119895

max 119903119894119895 minusmin 119903

119894119895

isin [0 1]

(6)

This standard transform formula can eliminate the influ-ence of fundamental units and unify CRI standardized valuesin the range of [0 1]


Table 1 Linguistic evaluation scale rule

2 3 4 5 5 6 7 9N Y Y Y Y YVL YL Y Y Y Y Y YRL Y Y Y YLL YM Y Y Y Y Y YLH YRH Y Y Y YH Y Y Y Y Y Y YVH YP Y Y Y Y Y

4 Classification Analysis of CRI

As mentioned in the CRI standardization procedure CRIfeatures with values in CRIDs can be transformed intocorresponding GCRI with cases in a structured model If theGCRI is regarded as the centroid the CRI features can beclassified by their dissimilarity with GCRI cases Because it ispossible to change the selection of GCRI flexibly according tospecific conditions such as technical abilities or informationvariation the classification can achieve diversification

The classification analysis of CRI performs the followingfive steps

(A) If there are 119899 CRIDs in accordance with the GCRIfeature set 119865

1 1198652 119865

119898 there will be 119898 selected

CRI features to constitute an 119899 times 119898matrix119883

119883 =

[[[[[[

[

11990911

11990912


11990921

11990922


1199091198991

1199091198992

sdot sdot sdot 119909119899119898

]]]]]]

]

(7)

(B) Thedissimilarities amongCRIDs are calculated by theEuclidean distance which is a matrix119863

119863 =

[[[[[[[[[

[

0

11988921

0

11988931

11988932

0

d

1198891198991

1198891198992

sdot sdot sdot 119889119899119898

0

]]]]]]]]]

]

(8)

(C) In (8) if 119889119894119895= 0 CRID 119894 and CRID 119895 will be merged

such that the dimensionality of matrix 119863 is reducedAn ℎ times 119898 (ℎ le 119899)matrix is shown as 119884

119884 =

[[[[[[

[

11991011

11991012


11991021

11991022


119910ℎ1

119910ℎ2

sdot sdot sdot 119910ℎ119898

]]]]]]

]

(9)

Table 2 Quantitative analysis of scaled feature variables

CRI scaled features Real values Standardized values

CPU 2-core 04-core 1

Frequency10G 012 G 0415 G 1

Memory RAM 512M 01536M 1

Dimension43 IN 045 IN 0547 IN 1

Screen resolution800 times 480 0854 times 480 019960 times 540 1

Table 3 Quantitative analysis of nominal binary feature variables

CRI scaledfeatures States of values Standardized values

Batteryreplacement Replaceable Irreplaceable 1 0

Message toenterprise Not null Null 1 0

(D) In the GCRI feature set 1198651 1198652 119865

119898 (119898 isin forall119873)

because each featuremay have no fewer than one casethe dissimilarity measurement formula between 119865

119896

and 119910sdot119896(119896 = 1 2 119898) is

119863lowast= min [1003816100381610038161003816119910sdot119896 minus 119865

1198961003816100381610038161003816] (10)

(E) In (10) if 119863lowast = 0 the feature value will be markedand its corresponding CRID will be submitted andrecorded The classification results depend on thenumber of119863lowast = 0 in the CRID

5 Case Study

51 CRI Quantitative Standardization for a Smart Phone TheCRI scaled features for a smart phone have real values withfundamental units some of which are chosen for the demon-stration of quantitative standardization The data is initiallysubjected to the process of quantitative analysis and standardtransformation described in Section 3 and the resultingstandardized values are listed in Table 2 Similarly the stan-dardized values of CRI nominal binary features are shown inTable 3

The CRI nominal multiple features for a smart phonehave no fewer than two states for example depth and colorBecause these states are not expressed by quantitative valuesthey have to be transformed by fuzzy mathematics and lin-guistic representation to obtain quantitative values which aresubsequently changed into standardized values The processof quantitative standardization is presented in Tables 4 and 5

In Table 4 three states of depth are transformed into atrapezoidal distribution by the linguistic evaluation scale rule


Table 4 Quantitative analysis of nominal multiple feature variables depth

Feature (depth) Trapezoidal distribution Quantitative values Standardized valuesUltrathin (00 00 09 09) 045 0Thin (09 09 10 10) 095 078Ordinary (10 10 12 12) 110 1

Table 5 Quantitative analysis of nominal multiple feature variables color

Feature (color) Fuzzy set vector Quantitative values Standardized valuesBlack (000 000 000) 000 0White (100 100 100) 173 1Grey (050 050 050) 087 05Light grey (083 083 083) 144 083Dark grey (066 066 066) 114 066Dim grey (041 041 041) 071 041Red (100 000 000) 100 058

Table 6 GCRI feature set and cases

Feature set Standardized values

Frequency0041

Color 01

Depth0

0781

Dimension0051

Screen resolution0

0191

Message to enterprise 0

and fuzzy membership distribution Then the quantitativevalues of defuzzification are calculated byChenrsquosmodel Afterperforming a standard transformation the standardized val-ues of the three depth states are 0 078 and 1 respectively

Table 5 contains the results of the transformation of sevenstates of color into a fuzzy set vector by a color universe basedon RGB [30] The quantitative values and the standardizedvalues are calculated by a defuzzification formula of a fuzzyset vector

52 CRI Classification Analysis for a Smart Phone If theGCRI feature set and their cases (set of standardized values)are selected as indicated in Table 6 these cases will producethe quantitative standardized feature values of 30 CRIDslisted in Table 7 which constitutes a 30 times 6 matrix In accor-dance with the dissimilarity measurement the dimensionreduction matrix is a 22 times 6 matrix

Finally the values in Table 7 were processed in termsof the dissimilarity measurement formula for selected GCRI

0

02

04

06

08

1

12

2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 221

Dlowast

ColorD

lowastMessage

Figure 3 Dissimilar distribution of CRIDs

GCRI Product family Smart phone prototype

Figure 4 Product configuration with GCRI and product family

cases and their corresponding feature values and the resultsare listed in Table 8The dissimilarity distribution is shown inFigure 3

According to results of dissimilarity with GCRI casesthose matching CRIDs are 1 5 8 10 14 15 17 19 20 21 2223 24 26 29 which means that they are compatible withproduct family model Their product configuration schemecan be generated Contrarily the remaining CRIDs and their


No matching 4500

Matching 5500

Six specific 070Five specific 620

Four specific 850

Three specific 1280

Two specific 2680

One specific 4500

Global analysis

10000

Requirements5500 4500

Not512

2015Matchingmatching

Figure 5 Global analysis diagram

Table 7 Quantitative standardized feature values for smart phone

CRID Frequency Color Depth Dimension SR Message1 04 0 0 05 019 02 04 1 078 05 019 13 04 087 078 05 019 14 04 0 078 05 019 15 04 0 078 05 019 06 04 05 078 05 019 17 04 05 078 05 019 08 04 1 078 05 019 09 04 1 078 05 019 110 04 1 078 05 019 011 0 1 078 1 0 112 0 1 078 1 0 113 0 087 078 1 0 114 0 0 078 1 0 015 0 0 078 1 0 016 0 0 0 1 0 117 0 0 0 1 019 018 0 0 0 1 019 119 0 0 0 1 019 020 0 0 0 0 1 021 1 0 0 0 1 022 1 0 0 0 1 023 1 0 0 0 1 024 1 0 0 0 1 025 1 05 0 0 1 026 1 1 0 0 1 027 1 05 0 1 1 128 1 1 0 1 1 129 1 1 0 1 1 030 1 05 0 1 1 0

features must be divided and marked by the dissimilarity Inview of the overall situation the classification analysis deriveswhat CRIDs and which CRI can meet product configurationFrom the microscopic view the classification analysis alsoderives which GCRI feature is easy to be challenged or

ignored Thus in terms of those challenged GCRI featuresthe product family model would be considered to makeappropriate adjustments for better product configurationFurthermore because the GCRI is able to be renovated it ispossible to achieve a flexibly classification if it is the centroid

53 Further Discussion Production in enterprises is largelyoriented towards CRI Because original CRI is hardly utilizedformodeling product family GCRI is widely adopted in orderto transform CRI into a structure model This structuredmodel composed of GCRI refers to product family modelBased on the GCRI and the product family the modularproduct configuration is able to select appropriate modulesand manufacture desirable products Take the case of a smartphone Figure 4 presents product configuration with GCRIand its product family Thus using GCRI to distinguish CRIcan effectively assist in product configuration

In CRI classification analysis for the smart phone thedissimilar distribution indicates what CRIDs can meet prod-uct configuration The global analysis diagram shown inFigure 5 affords a proportion of dissimilarity with a GCRI setin all CRIDs and that with GCRI cases in each CRID Theenterprise can determine the processing level of each CRIDaccording to the proportion of dissimilarity between GCRIcases whereas CRIDs with a higher proportion often requiremore complex treatment

Likewise the CRI classification analysis discovers whichGCRI feature is easy to be challenged or ignored The featureanalysis diagram shown in Figure 6 exposes the statistics offeatures that differ from those in theGCRI cases Popular CRIfeatures such as the color feature are clearly shown More-over enterprises are able to select those elements they wish toinclude in the GCRI feature set by either adding or removingfeatures presented in this analysis result These results areexpected to be useful for product family renewal

6 Conclusions

This paper proposes a classification approach for realizingCRI quantitative analysis aimed at supporting product familyadaptation in enterprises The CRI classification analysis notonly considers the scalability of classification results but alsoregards subsequent application to product configuration


Table 8 Dissimilarity results

CRID 119863lowast

Frequency 119863lowast

Color 119863lowast

Depth 119863lowast

Dimension 119863lowast

SR 119863lowast

Message

1 1 0 0 0 0 0 02 2 amp 9 0 0 0 0 0 13 3 0 013 0 0 0 14 4 0 0 0 0 0 15 5 0 0 0 0 0 06 6 0 05 0 0 0 17 7 0 05 0 0 0 08 8 amp 10 0 0 0 0 0 09 11 amp 12 0 0 0 0 0 110 13 0 013 0 0 0 111 14 amp 15 0 0 0 0 0 012 16 0 0 0 0 0 113 17 amp 19 0 0 0 0 0 014 18 0 0 0 0 0 115 20 0 0 0 0 0 016 21 amp 22 amp 23 amp 24 0 0 0 0 0 017 25 0 05 0 0 0 018 26 0 0 0 0 0 019 27 0 05 0 0 0 120 28 0 0 0 0 0 121 29 0 0 0 0 0 022 30 0 05 0 0 0 0

Feature amount

0

250

500

750

Col

or

Dep

th

Freq

uenc

y

Dim

ensio

n

Scre

en re

solu

tion

Mes

sage

to en

terp

rise

Figure 6 Feature analysis diagram

At the technical level considering that CRI feature valuesconsist of multiple data types such as numbers and wordsa quantitative analysis based on fuzzy mathematics andlinguistic representation is presentedThis analysis is capableof revealing not only the differences between the CRI featuresof a product but also the differences among the states ineach CRI feature thereby avoiding shortcomings such asincomplete expression of states and meaningless assignmentof features Furthermore the dissimilarity amongCRI featurevalues is measured by utilizing a standard transformation foreliminating the influence of different fundamental units Anassociation between the classification analysis and productconfiguration is achieved by using a flexible classification

based on the fact that the selective GCRI is regarded as thecentroid Therefore the determination of the classificationresults is no longer an isolated operation instead it deriveswhich CRI can meet product configuration and which GCRIfeature can assist in improving product family

In engineering practice classification analysis enablesCRI to be quantified recognized information which will becompatible with other management systems in enterprisessuch as ERP or PDM It is helpful for the final productrapid development and intelligent configuration Meanwhileusing GCRI to discover the specific features enterprises candetermine market positioning of their future product andpredict the corresponding product family model Althoughthe proposed approach is demonstrated by analyzing selectedfeatures of smart phones to verify the feasibility and effective-ness it can be extended to other consumer electronics

As futurework wewill consider the study of a new frame-work that enables CRI classification analysis to deal with BigData

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This research is funded and supported by the Natural ScienceFoundation of Hubei Province China (2015CFA115) andScience and Technology Support Program ofHubei Province


China (2015BAA058)The authors express their most sincereappreciation to Professor Quan LIU who provided an exper-imental platform together with valuable advice The authorsare also grateful for assistance received from Professor Qing-song AI and Professor Ping LOU regarding revising thepaper

References

[1] Q L Xu R J Jiao X YangMGHelanderHMKhalid andOAnders ldquoCustomer requirement analysis based on an analyticalKano modelrdquo in Proceedings of the IEEE International Confer-ence on Industrial Engineering andEngineeringManagement pp1287ndash1291 Singapore December 2007

[2] L Xie and Z Li ldquoA customer requirements ratingmethod basedon fuzzy kano modelrdquo in Future Control and Automation vol 173 of Lecture Notes in Electrical Engineering pp 141ndash146Springer Berlin Germany 2012

[3] M A Shafia and S Abdollahzadeh ldquoIntegrating fuzzy kano andfuzzy TOPSIS for classification of functional requirements innational standardization systemrdquo Arabian Journal for Scienceand Engineering vol 39 no 8 pp 6555ndash6565 2014

[4] Z Lingyun Research on the Key Technology of CustomerRequirements Processing ofDesign forMass Customization HefeiUniversity of Technology Hefei China 2008

[5] S Li Y Liu J Wang and H Zeng ldquoAn intelligent interactiveapproach for assembly process planning based on hierarchicalclassification of partsrdquo The International Journal of AdvancedManufacturing Technology vol 70 no 9ndash12 pp 1903ndash1914 2014

[6] Y G Jing B Dan S Peng and L F Guo ldquoIntelligent mappingof semi-structured customer needs for web-based productcustomizationrdquoAdvancedMaterials Research vol 201 pp 1496ndash1499 2011

[7] Z Xiao Q Liu and Q S Ai ldquoAcquisition and analysis for cus-tomer requirement informationrdquo Advanced Materials Researchvol 139ndash141 pp 1468ndash1471 2010

[8] China Internet Network Information Center (CNNIC)ldquoResearch Report of Chinese Online Shopping Marketrdquo 2013httpwwwcnniccn

[9] Y Wang and M M Tseng ldquoIntegrating comprehensive cus-tomer requirements into product designrdquo CIRP AnnalsmdashMan-ufacturing Technology vol 60 no 1 pp 175ndash178 2011

[10] Y Wang and M M Tseng ldquoIdentifying emerging customerrequirements in an early design stage by applying bayes factor-based sequential analysisrdquo IEEE Transactions on EngineeringManagement vol 61 no 1 pp 129ndash137 2014

[11] S Negash T Ryan and M Igbaria ldquoQuality and effectivenessin web-based customer support systemsrdquo Information amp Man-agement vol 40 no 8 pp 757ndash768 2003

[12] Y Tan and G Wei Theory and Method for Modular ProductIntelligent Configuration Southwest Jiaotong University Press2011

[13] J Wang Research on Identification and Description Approachesto the Information about Customer Needs for Mass Customiza-tion Chongqing University Chongqing China 2009

[14] Y Wang Information Extraction from Clinical Notes Universityof Sydney Sydney Australia 2009

[15] G Salton A Wong and C S Yang ldquoA vector space model forautomatic indexingrdquo Communications of the ACM vol 18 no11 pp 613ndash620 1975

[16] J Han M Kamber and J Pei Data Mining Concepts and Tech-niques Morgan Kaufmann San Francisco Calif USA 2006

[17] L Jing M K Ng and J Z Huang ldquoKnowledge-based vectorspace model for text clusteringrdquo Knowledge and InformationSystems vol 25 no 1 pp 35ndash55 2010

[18] A Strehl J Ghosh and R Mooney ldquoImpact of similarity mea-sures onweb-page clusteringrdquo inProceedings of theWorkshop onArtificial Intelligence for Web Search (AAAI rsquo00) pp 58ndash64 July2000

[19] J Xie and C Liu Methods and Applications of Fuzzy Mathe-matics Huazhong University of Science and Technology PressWuhan China 2006

[20] X Bi X Zhao G Wang Z Zhang and S Chen ldquoDistributedlearning over massive XML documents in ELM feature spacerdquoMathematical Problems in Engineering vol 2015 Article ID923097 13 pages 2015

[21] A S Fathima D Manimegalai and N Hundewale ldquoA reviewof data mining classification techniques applied for diagnosisand prognosis of the arbovirus-denguerdquo International Journalof Computer Science Issues vol 8 no 6 p 322 2011

[22] K Wagstaff C Cardie S Rogers and S Schrodl ldquoConstrainedk-means clustering with background knowledgerdquo in Proceed-ings of the 18th International Conference on Machine Learning(ICML rsquo01) vol 1 pp 577ndash584 Williamstown Mass USA June2001

[23] T Kanungo D M Mount N S Netanyahu C D Piatko RSilverman and A Y Wu ldquoAn efficient k-means clustering algo-rithm analysis and implementationrdquo IEEE Transactions on Pat-ternAnalysis andMachine Intelligence vol 24 no 7 pp 881ndash8922002

[24] Wikipedia Euclidean Distance 2015 httpsenwikipediaorgwikiEuclidean distance

[25] Y Chen M Rege M Dong and J Hua ldquoNon-negative matrixfactorization for semi-supervised data clusteringrdquo Knowledgeand Information Systems vol 17 no 3 pp 355ndash379 2008

[26] L A Zadeh Fuzzy Sets and Fuzzy Information GranulationTheory Beijing Normal University Publishing Group BeijingChina 2005

[27] S-M Chen ldquoFuzzy group decision making for evaluating therate of aggregative risk in software developmentrdquo Fuzzy Sets andSystems vol 118 no 1 pp 75ndash88 2001

[28] F Herrera and LMartınez ldquoA 2-tuple fuzzy linguistic represen-tation model for computing with wordsrdquo IEEE Transactions onFuzzy Systems vol 8 no 6 pp 746ndash752 2000

[29] F Herrera and L Martınez ldquoA model based on linguistic 2-tuples for dealing with multigranular hierarchical linguisticcontexts in multi-expert decision-makingrdquo IEEE Transactionson Systems Man and Cybernetics Part B Cybernetics vol 31no 2 pp 227ndash234 2001

[30] Z Zhou Z Xiao Q Liu and Q Ai ldquoAn analytical approachto customer requirement information processingrdquo EnterpriseInformation Systems vol 7 no 4 pp 543ndash557 2013

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MathematicsJournal of


Mathematical Problems in Engineering

Hindawi Publishing Corporationhttpwwwhindawicom

Differential EquationsInternational Journal of

Volume 2014

Applied MathematicsJournal of


Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of


Mathematical PhysicsAdvances in

Complex AnalysisJournal of


OptimizationJournal of


CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of


Operations ResearchAdvances in

Journal of


Function Spaces

Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of Mathematics and Mathematical Sciences


The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Algebra

Discrete Dynamics in Nature and Society



Decision SciencesAdvances in

Discrete MathematicsJournal of


Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Stochastic AnalysisInternational Journal of


In conclusion traditional classificationmethods focus oncustomer requirements rather than on product configurationsuch that the classification results are neither extendable norscalable The processes of CRI classification and productconfiguration can occur independently of one another Forinstance customersrsquo preferences are able to complete trans-formation into product design [9 10] without performingCRI classification although CRI classification results areuseful for determining them Additionally the lack of mathe-matical analysis of CRImeans that it is difficult to use the clas-sification results directly for product configuration purposesTherefore the modern approach to CRI classification has notbeen a simple issue of dividing the information into groupsinstead it has to consider (i) how to analyze and express CRIwith mathematical methods and (ii) how to classify CRI forthe purpose of product configuration

In this paper we introduce a classification analysis basedon quantitative standardization Section 2 describes a math-ematical model to analyze CRI in terms of product fami-lies Section 3 provides details as to how to transform CRIinto quantitative standards On the basis of Sections 2 and3 Section 4 presents the classification method Finally inSection 5 a case is demonstrated to confirm the feasibility andeffectiveness of the proposed method

2 Background Review

21 CRI Structured Procedure The increasing application ofe-commerce has been transforming the acquisition methodof choice from offline to online Online CRI acquisitionmainly depends on theWeb including the advantages it offersin terms of efficiency affordability and convenience [11]Online CRI can be submitted in the form of XMLdocumentswhich can containmultiple kinds of data and can adapt to thedynamic development of CRI AnXMLdocument containingCRI is referred to as a customer requirement informationdocument (CRID) Because of the XML framework eachCRID can tag various CRI features of each customer how-ever these features contain multiple data types character-ized by fuzziness concealment and similarity such thatit is difficult to correctly identify the information for use inthe process of product configuration Thus it is necessary totransform CRI into a structured model corresponding to theproduct family model to enable the information to be trans-lated into product development for manufacturing Researchon the structure model of the product family has led to theconstruction of general customer requirement information(GCRI) [12] which is able to abstract a series of similarCRI featureswhose personalized features are distinguished byspecific values of cases [13] Figure 1 illustrates the standard-ization procedure in which CRI is transformed into GCRIwhich means that features with values in CRID categoriescan be transformed into the corresponding GCRI classes in astructured model

22 CRI Document Model In this model the CRI is submit-ted in the form of CRID which is a document based on adocument representation model capable of enlightening the

CRID1CRID2

CRID3CRI structure model

C11 C12 C21 C22 C23 C31 C32

GCRI (F1) GCRI (F2) GCRI (F3)

middot middot middot

Figure 1 Summary of the standardization procedure

CRID representation Because a document is composed ofwords theword is themostwidely used unit of information indocumentmodeling [14] namely a document representationmodel can be established by using the characteristics of wordsand is implemented by the Vector Space Model (VSM) TheVSM [15] is a vector model utilizing the extracted character-istics of words from a document in a Euclidean space [16] Aword characteristic corresponds to a separate term If a termoccurs in the document its value in the vector is nonzero [17]The main idea of VSM is as follows let 119863 = [119863

1 1198632 119863

119889]

be a document set with each document 119863119894(119894 = 1 2 119889)

represented by a set of terms 119879 = [1198791 1198792 119879

119905] any

119879119895(119895 = 1 2 119905) corresponds to one dimension in the

VSM such that 119863 can be a 119889 times 119905 matrix which means thatdocuments can be mapped to a point in the VSM and theirsimilarity can be calculated by distances To date the VSM isthe most efficient and useful document representation modelbecause it transforms the similarity between two documentsinto the similarity between two vectors [18]

Thus the representation model of the CRID is a one-dimensional vector containing the values of the CRI featuresIf the number of CRID tags is 119899 and the CRI features areindependent of each other anyCRID119863

119894can be represented as

a one-dimensional set [1199031198941 1199031198942 119903

119894119899] On the basis of this if

the number of CRIDs is119898 and each CRID defines 119899 featuresthere will be an119898 times 119899matrix shown as follows

119877 =

[[[[[[

[

11990311

11990312


11990321

11990322


1199031198981

1199031198982

sdot sdot sdot 119903119898119899

]]]]]]

]

(1)

In the matrix 119877 119903119894119895(119894 = 1 2 119898 119895 = 1 2 119899)

is a 119895 feature value submitted by one customer 119894 Howeveran existing problem of CRI features is that the informationmay occur as a combination of multiple data types such asnumbers andwordsThis issue would have to be addressed byanalyzing the similarity of these values to introduce quantita-tive standardization of CRI such that 119903

119894119895in matrix 119877 could be

changed into a uniform fundamental unit [19]

23 Classification and Clustering Classification and clus-tering are two major methods for information analysisespecially data in XML documents [20] The aim of theclassification is to build a classifier based on some cases withsome attributes to describe the objects or one attribute todescribe the group of the objectsThen the classifier is used to












119860

119880 rarr [0 1] 119906 997891rarr

120583119860


119860



be expressed as

119860= (120583119860

(1199061) 120583119860

(1199062) 120583

119860

(119906119899)) (2)


119860 = radic120583119860

(1199061)2

+ 120583119860

(1199062)2

+ sdot sdot sdot + 120583119860

(119906119899)2

(3)


is described by




120583119872

(119906) =

119906 minus 119897

1198981minus 119897

119897 le 119906 le 1198981

1 1198981le 119906 le 119898

2

119906 minus 119899

1198982minus 119899

1198982le 119906 le 119899

0 others

(4)


is

119872 =119897 + 1198981+ 1198982+ 119899

4 (5)



119903119894119895=

119903119894119895minus 119903119895

radic(1 (119898 minus 1))sum119898

119894=1(119903119894119895minus 119903119895)2

isin [minus1 1]

119903lowast

119894119895=

119903119894119895minusmin 119903

119894119895

max 119903119894119895 minusmin 119903

119894119895

isin [0 1]

(6)









1 1198652 119865



119883 =

[[[[[[

[

11990911

11990912


11990921

11990922


1199091198991

1199091198992

sdot sdot sdot 119909119899119898

]]]]]]

]

(7)


119863 =

[[[[[[[[[

[

0

11988921

0

11988931

11988932

0

d

1198891198991

1198891198992

sdot sdot sdot 119889119899119898

0

]]]]]]]]]

]

(8)



119884 =

[[[[[[

[

11991011

11991012


11991021

11991022


119910ℎ1

119910ℎ2


]]]]]]

]

(9)













119898 (119898 isin forall119873)


119896

and 119910sdot119896(119896 = 1 2 119898) is


1198961003816100381610038161003816] (10)


5 Case Study











Frequency0041

Color 01

Depth0

0781

Dimension0051

Screen resolution0

0191






0

02

04

06

08

1

12

2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 221

Dlowast

ColorD

lowastMessage







No matching 4500

Matching 5500


Four specific 850

Three specific 1280

Two specific 2680

One specific 4500

Global analysis

10000


Not512










6 Conclusions




CRID 119863lowast


Color 119863lowast

Depth 119863lowast


SR 119863lowast

Message


Feature amount

0

250

500

750

Col

or

Dep

th

Freq

uenc

y

Dim

ensio

n

Scre

en re

solu

tion

Mes

sage

to en

terp

rise








Acknowledgments




References






































Volume 2014




Journal of











Journal of


Function Spaces






Algebra




















119860

119880 rarr [0 1] 119906 997891rarr

120583119860


119860



be expressed as

119860= (120583119860

(1199061) 120583119860

(1199062) 120583

119860

(119906119899)) (2)


119860 = radic120583119860

(1199061)2

+ 120583119860

(1199062)2

+ sdot sdot sdot + 120583119860

(119906119899)2

(3)


is described by




120583119872

(119906) =

119906 minus 119897

1198981minus 119897

119897 le 119906 le 1198981

1 1198981le 119906 le 119898

2

119906 minus 119899

1198982minus 119899

1198982le 119906 le 119899

0 others

(4)


is

119872 =119897 + 1198981+ 1198982+ 119899

4 (5)



119903119894119895=

119903119894119895minus 119903119895

radic(1 (119898 minus 1))sum119898

119894=1(119903119894119895minus 119903119895)2

isin [minus1 1]

119903lowast

119894119895=

119903119894119895minusmin 119903

119894119895

max 119903119894119895 minusmin 119903

119894119895

isin [0 1]

(6)









1 1198652 119865



119883 =

[[[[[[

[

11990911

11990912


11990921

11990922


1199091198991

1199091198992

sdot sdot sdot 119909119899119898

]]]]]]

]

(7)


119863 =

[[[[[[[[[

[

0

11988921

0

11988931

11988932

0

d

1198891198991

1198891198992

sdot sdot sdot 119889119899119898

0

]]]]]]]]]

]

(8)



119884 =

[[[[[[

[

11991011

11991012


11991021

11991022


119910ℎ1

119910ℎ2


]]]]]]

]

(9)













119898 (119898 isin forall119873)


119896

and 119910sdot119896(119896 = 1 2 119898) is


1198961003816100381610038161003816] (10)


5 Case Study











Frequency0041

Color 01

Depth0

0781

Dimension0051

Screen resolution0

0191






0

02

04

06

08

1

12

2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 221

Dlowast

ColorD

lowastMessage







No matching 4500

Matching 5500


Four specific 850

Three specific 1280

Two specific 2680

One specific 4500

Global analysis

10000


Not512










6 Conclusions




CRID 119863lowast


Color 119863lowast

Depth 119863lowast


SR 119863lowast

Message


Feature amount

0

250

500

750

Col

or

Dep

th

Freq

uenc

y

Dim

ensio

n

Scre

en re

solu

tion

Mes

sage

to en

terp

rise








Acknowledgments




References






































Volume 2014




Journal of











Journal of


Function Spaces






Algebra
















1 1198652 119865



119883 =

[[[[[[

[

11990911

11990912


11990921

11990922


1199091198991

1199091198992

sdot sdot sdot 119909119899119898

]]]]]]

]

(7)


119863 =

[[[[[[[[[

[

0

11988921

0

11988931

11988932

0

d

1198891198991

1198891198992

sdot sdot sdot 119889119899119898

0

]]]]]]]]]

]

(8)



119884 =

[[[[[[

[

11991011

11991012


11991021

11991022


119910ℎ1

119910ℎ2


]]]]]]

]

(9)













119898 (119898 isin forall119873)


119896

and 119910sdot119896(119896 = 1 2 119898) is


1198961003816100381610038161003816] (10)


5 Case Study











Frequency0041

Color 01

Depth0

0781

Dimension0051

Screen resolution0

0191






0

02

04

06

08

1

12

2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 221

Dlowast

ColorD

lowastMessage







No matching 4500

Matching 5500


Four specific 850

Three specific 1280

Two specific 2680

One specific 4500

Global analysis

10000


Not512










6 Conclusions




CRID 119863lowast


Color 119863lowast

Depth 119863lowast


SR 119863lowast

Message


Feature amount

0

250

500

750

Col

or

Dep

th

Freq

uenc

y

Dim

ensio

n

Scre

en re

solu

tion

Mes

sage

to en

terp

rise








Acknowledgments




References






































Volume 2014




Journal of











Journal of


Function Spaces






Algebra
















Frequency0041

Color 01

Depth0

0781

Dimension0051

Screen resolution0

0191






0

02

04

06

08

1

12

2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 221

Dlowast

ColorD

lowastMessage







No matching 4500

Matching 5500


Four specific 850

Three specific 1280

Two specific 2680

One specific 4500

Global analysis

10000


Not512










6 Conclusions




CRID 119863lowast


Color 119863lowast

Depth 119863lowast


SR 119863lowast

Message


Feature amount

0

250

500

750

Col

or

Dep

th

Freq

uenc

y

Dim

ensio

n

Scre

en re

solu

tion

Mes

sage

to en

terp

rise








Acknowledgments




References






































Volume 2014




Journal of











Journal of


Function Spaces






Algebra










No matching 4500

Matching 5500


Four specific 850

Three specific 1280

Two specific 2680

One specific 4500

Global analysis

10000


Not512










6 Conclusions




CRID 119863lowast


Color 119863lowast

Depth 119863lowast


SR 119863lowast

Message


Feature amount

0

250

500

750

Col

or

Dep

th

Freq

uenc

y

Dim

ensio

n

Scre

en re

solu

tion

Mes

sage

to en

terp

rise








Acknowledgments




References






































Volume 2014




Journal of











Journal of


Function Spaces






Algebra











CRID 119863lowast


Color 119863lowast

Depth 119863lowast


SR 119863lowast

Message


Feature amount

0

250

500

750

Col

or

Dep

th

Freq

uenc

y

Dim

ensio

n

Scre

en re

solu

tion

Mes

sage

to en

terp

rise








Acknowledgments




References






































Volume 2014




Journal of











Journal of


Function Spaces






Algebra











References






































Volume 2014




Journal of











Journal of


Function Spaces






Algebra
















Volume 2014




Journal of











Journal of


Function Spaces






Algebra









research article a new classification analysis of customer...

Documents