文稿案例期末_第1頁
文稿案例期末_第2頁
文稿案例期末_第3頁
文稿案例期末_第4頁
文稿案例期末_第5頁
已閱讀5頁,還剩29頁未讀, 繼續(xù)免費(fèi)閱讀

下載本文檔

版權(quán)說明:本文檔由用戶提供并上傳,收益歸屬內(nèi)容提供方,若內(nèi)容存在侵權(quán),請(qǐng)進(jìn)行舉報(bào)或認(rèn)領(lǐng)

文檔簡(jiǎn)介

1、PowerPoint to panyNaresh MalhotraJohn HallMike ShawPeter OppenheimPowerPoint to panyChapter 19Segmenting ConsumersChapter ObjectivesAfter reading this chapter you should be able to:Describe concept and scope of cluster analysis and its importance in marketing researchDiscuss the statistics associate

2、d with cluster analysisOutline the procedures for conducting a cluster analysisDescribe the purpose and methods of evaluating the quality of clustering results and assessing reliability and validityDiscuss the applications of non-hierarchical clustering and clustering of variables Conduct a cluster

3、analysis for segmentation purposes using SPSS or similar softwareChapter OutlineBasic ConceptConducting Cluster AnalysisNon-hierarchical Clustering Clustering Variables Computer ApplicationsSummaryTopicBasic ConceptConducting Cluster AnalysisNon-hierarchical Clustering Clustering Variables Computer

4、ApplicationsSummaryBasic ConceptCluster analysis classifies objects or cases into relatively homogenous groups called “clusters”Ideally clusters should be distinctEach object is assigned to only one clusterVariable 2Variable 1Figure 19.1: Ideal ClusteringBasic ConceptIn a practical sense clustering

5、is more likely to be like thisBoundaries are not clear-cut and classification of some consumers into a group is more difficultCluster analysis and discriminant analysis but:Latter requires a priori knowledge of cluster membershipIn cluster analysis group membership suggested by the dataMarketing App

6、licationsMarket segmentation:Along benefit linesUnderstanding buyer behaviour:Form homogenous groups and then analyse each group separatelyIdentifying new product opportunities:Cluster brands and product give rise to competitive sets and opportunitiesSelecting test markets:Select comparable cities f

7、or test strategiesReduce data:Conduct analysis on more manageable data setsTopicBasic ConceptConducting Cluster AnalysisNon-hierarchical Clustering Clustering Variables Computer ApplicationsSummaryConducting Cluster AnalysisFormulate the ProblemSelect variables for basis of clustering:In a marketing

8、 sense describe similarity between objectsPast research, theory of hypotheses being consideredFollowing illustration is of clustering consumers based on attitudes toward shoppingExpress extent of agreement on number of statements:agree (7) disagree (1) Formulate the ProblemExamples of statements:V1

9、- shopping is funV3 - I combine shopping with eating outV6 - I dont care about shoppingTable 19.1: Attitudinal Data for ClusteringNB: clustering normally conducted on samples of 100 plusSelect a Distance or Similarity MeasureMost common approach is to measure similarity distance between pairs of obj

10、ectsEuclidean distance:Square root of the sum of the squared differences in values for each variableIf variables are measured in different units then need to standardise rescale to mean = 0 and SD of unity before clusteringStandardisation can reduce differences between groups on the best discriminat

11、ing variablesDesirables to eliminate outliersVisits pm, $ spent, % spent at a shop (loyalty)Figure 19.4: Clustering ProceduresSelect a Clustering ProcedureHierarchical clustering - develops a tree-like structure:AgglomerativeEach object starts out in a separate clusterClusters formed by grouping obj

12、ects into bigger and bigger clustersStops when objects are members of a single clusterDivisiveAll objects in a single clusterDivided until each object in a separate clusterAgglomerative common in marketing researchVarious methods availableSelect a Clustering Procedure (continued)Linkage methods clus

13、ters objects based on distance between them:Single linkageMinimum distanceComplete linkageMaximum distanceAverage linkageMaximum distancePreferred as it uses information on all pairsSelect a Clustering Procedure (continued)Variance methods attempt to minimise within-cluster variance:WardsSquared Euc

14、lidean distance to the cluster means is minimisedCentroidDistance between two clusters is the distance between their two centroids (means for all the variables)Select a Clustering ProcedureNon-Hierarchical Clustering first assigns or determines a cluster centre and then groups all objects within a p

15、respecified threshold value from the centre:Sequential thresholdA cluster centre is selected and all objects within a prespecified threshold value from the centre are grouped togetherParallel thresholdSpecifies several cluster centres at onceAll objects with threshold grouped togetherOptimising part

16、itioningAllows for later re-assignment of objects to clusters to optimise an overall criterionSelect a Clustering ProcedureNon-hierarchical disadvantages:Cluster numbers must be prespecifiedSelection of cluster centres is arbitraryChoice of clustering method and choice of a distance measure are inte

17、rrelatedReturning to the shopping illustration:Dendrogram shows which clusters are joined together but sequence for early clusters is difficult to tellIn last two stages, the distances at which the clusters are being formed, are largeSee pp. 736-8 and Appendix 19B for details of this illustrationUse

18、ful for deciding number of clustersDendrogramDecide on the Number of ClustersGuidelines:Theoretical, conceptual or practical considerationsIn hierarchical clustering, distances used as criteriaIn non-hierarchical, ratio of within-group variance to between group variance plotted against cluster numbe

19、rs appropriate number at the elbowRelative cluster sizes should be meaningfulInterpret and Profile the ClustersInvolves examining cluster centroids:Mean values of the objects contained in the cluster on each variableFinal cluster centre aids naming the clusterCluster 3 “fun-loving and concerned shop

20、pers”Cluster 2 “apathetic shoppers”Cluster 1 “economical shoppers”Save cluster membership as a new variable so that you can build a profile:DemographicPsychographicProduct usage, etcSee pp. 738-9 for complete details of the shopping attitude clusteringBetter targetingSee pp. 740-1 for an illustratio

21、n of hierarchical clusteringAssess Reliability and ValidityProcedures:Use same data but different distance measures and compare stability of resultsUse different clustering methods and compareRandom split data in halfPerform clusteringCompare centroidsDelete variables randomlyPerform clusteringCompa

22、re results against those of full data setIn non-hierarchicalSolution may depend on order of cases, soDo multiple runs (different case orders) until stableTopicBasic ConceptConducting Cluster AnalysisNon-hierarchical Clustering Clustering Variables Computer ApplicationsSummaryNon-hierarchical Cluster

23、ingUsing data from Table 19.1 and K-means cluster procedure in SPSS (Appendix 19B):Results in Figure 19.8Initial cluster centres are the values of the first three casesFinal cluster centres represent the variable means for the cases and the final clustersUnivariate F test for each clustering variabl

24、e is presentedCases systematically assigned to clusters to maximise differences on clustering variablesResulting probabilities should not be interpreted as testing the null hypothesis of no differencesSee Example 19.3 pp. 741-2 for a non-hierarchical exampleFigure 19.8: Non-hierarchical ClusteringFi

25、gure 19.8: Non-hierarchical Clustering (continued)TopicBasic ConceptConducting Cluster AnalysisNon-hierarchical Clustering Clustering Variables Computer ApplicationsSummaryClustering VariablesAids identification of unique variablesReduces the number of variablesAssociated with each cluster is a linear combination of the variables called the “cluster component”Large set of variables can be replaced by a set of cluster components with little loss of informationCluster

溫馨提示

  • 1. 本站所有資源如無特殊說明,都需要本地電腦安裝OFFICE2007和PDF閱讀器。圖紙軟件為CAD,CAXA,PROE,UG,SolidWorks等.壓縮文件請(qǐng)下載最新的WinRAR軟件解壓。
  • 2. 本站的文檔不包含任何第三方提供的附件圖紙等,如果需要附件,請(qǐng)聯(lián)系上傳者。文件的所有權(quán)益歸上傳用戶所有。
  • 3. 本站RAR壓縮包中若帶圖紙,網(wǎng)頁內(nèi)容里面會(huì)有圖紙預(yù)覽,若沒有圖紙預(yù)覽就沒有圖紙。
  • 4. 未經(jīng)權(quán)益所有人同意不得將文件中的內(nèi)容挪作商業(yè)或盈利用途。
  • 5. 人人文庫(kù)網(wǎng)僅提供信息存儲(chǔ)空間,僅對(duì)用戶上傳內(nèi)容的表現(xiàn)方式做保護(hù)處理,對(duì)用戶上傳分享的文檔內(nèi)容本身不做任何修改或編輯,并不能對(duì)任何下載內(nèi)容負(fù)責(zé)。
  • 6. 下載文件中如有侵權(quán)或不適當(dāng)內(nèi)容,請(qǐng)與我們聯(lián)系,我們立即糾正。
  • 7. 本站不保證下載資源的準(zhǔn)確性、安全性和完整性, 同時(shí)也不承擔(dān)用戶因使用這些下載資源對(duì)自己和他人造成任何形式的傷害或損失。

評(píng)論

0/150

提交評(píng)論