Implementation of clustering techniques for segmentation of Mozambican cassava suppliers

dc.contributor.advisorGrobler, Jacomineen_ZA
dc.contributor.authorMatshabaphala, Ntebaleng Sharonen_ZA
dc.contributor.otherStellenbosch University. Faculty of Engineering. Dept. of Industrial Engineering.en_ZA
dc.date.accessioned2021-02-08T18:42:08Z
dc.date.accessioned2021-04-21T14:38:25Z
dc.date.available2021-02-08T18:42:08Z
dc.date.available2021-04-21T14:38:25Z
dc.date.issued2021-03
dc.descriptionThesis (MEng)--Stellenbosch University, 2021.en_ZA
dc.description.abstractENGLISH ABSTRACT: Although an organisation generally accumulates many suppliers in the course of doing business,some of these suppliers are of little or no importance to the organisation beyond fulfilling a simple order transaction, while other suppliers play a strategic role in the success of an organisation. The decision to invest in supplier relationships is a major step for an organisation, especially because the value gained from interacting in a supply network rests on the principle of prioritising the right suppliers. The segmentation of suppliers plays a significant role in supplier relationship management. Not only does it offer an effective method of assessing suppliers, but it also provides a resource-efficient decision methodology that specifies appropriate relation-ships and governance structures for each segment. In this thesis, three techniques are applied for clustering cassava suppliers in Mozambique.Over 3 000 smallholder farmers supply cassava to a for-profit social enterprise called Dadtco Philafrica. Dadtco Philafrica needs an effective supplier segmentation method to gain insight into how it should direct its resources to where they will have the greatest impact. Thek-means algorithm, agglomerative hierarchical clustering (AHC), and self-organising maps(SOM) with Ward clustering were applied to a real-world case study. Extensive algorithm pa-rameter tuning was conducted in order to ascertain good parameter values for each clustering technique. Performance of the algorithms was evaluated and compared using intra-cluster and inter-cluster distances, and the best performing algorithm, in the context of the case study,was selected. The SOM with Ward clustering outperformed thek-means and AHC, and its results were used to conduct a detailed cluster analysis. The insights gained from the cluster analysis were used to provide recommendations and to suggest suitable intervention strategies to manage each segment of suppliers. The encouraging results of these algorithms showed that clustering techniques can be utilised effectively in segmenting suppliers. The proposed method offers users the basis of a suppliers egmentation system that is more efficient. A user can simply rerun the algorithm using the latest data, to check for suppliers who have moved to a different cluster and to determine cluster allocation of new suppliers. This method relies primarily on historical data to segment suppliers; therefore, it provides an organisation with data-based insight regarding its supplybase.en_ZA
dc.description.abstractAFRIKAANSE OPSOMMING: Alhoewel ’n organisasie oor die algemeen heelwat verskaffers deur die loop van sake versamel,is sommige van hierdie verskaffers van min of geen belang vir die organisasie buiten om ’neenvoudige besteltransaksie uit te voer, terwyl ander verskaffers ’n strategiese rol speel in diesukses van ’n organisasie. Die besluit om in verskaffersverhoudinge te belˆe, is ’n belangrike stap vir ’n organisasie, veralomdat die waarde wat uit die interaksie in ’n verskaffingsnetwerk verkry word, berus op diebeginsel van prioritisering van die regte verskaffers. Die segmentering van verskaffers speel ’nbelangrike rol in die bestuur van verskafferverhoudinge. Segmentering bied nie net ’n effektiewemetode om verskaffers te evalueer nie, maar ook ’n hulpbroneffektiewe besluitnemingsmetodolo-gie wat toepaslike verhoudings en bestuurstrukture vir elke segment spesifiseer.In hierdie tesis word drie tegnieke toegepas vir die groepering van kassava-verskaffers in Mosam-biek. Meer as 3 000 kleinboere lewer kassava aan ’n winsgewende maatskaplike ondernemingmet die naam Dadtco Philafrica. Dadtco Philafrica benodig ’n effektiewe verskaffersegmenter-ingsmetode om insig te bekom oor hoe sy hulpbronne aangewend moet word om die grootsteimpak te maak. Diek-gemiddelde groepering algoritme, agglomeratiewe hi ̈erargiese groepering (AHC) en selfor-ganiserende afbeelding (SOM) met ‘Ward’ groepering is toegepas op ’n werklike gevallestudie.Omvattende instelling van algoritme-parameters is uitgevoer om goeie parameter waardes virelke groeperingstegniek te bepaal. Die uitvoering van die algoritmes is ge ̈evalueer en vergelykten opsigte van intra-groep en inter-groep afstande, en die beste presterende algoritme, in diekonteks van die gevallestudie, is gekies. Die groepering van die SOM met ‘Ward’ groeperinghet beter gevaar as diek-gemiddelde groepering algoritme en AHC, en die resultate daarvan isgebruik om ’n gedetailleerde groepontleding uit te voer. Die insigte wat uit die groepontledingverkry is, is gebruik om aanbevelings te gee en geskikte intervensiestrategie ̈e voor te stel omelke segment van verskaffers te bestuur.Die bemoedigende resultate van hierdie algoritmes het getoon dat groeperingstegnieke effektiefin verskaffersegmentering gebruik kan word. Die voorgestelde metode bied gebruikers die basisvan ’n verskaffersegmenteringsstelsel wat meer doeltreffend is. ’n Gebruiker kan eenvoudigdie groepontleding oordoen deur die nuutste data te gebruik om verskaffers wat na ’n andergroep beweeg het te identifiseer, en om die groepering van nuwe verskaffers te bepaal. Hierdiemetode maak hoofsaaklik staat op historiese data vir verskaffersegmentering; daarom bied dit’n organisasie data-gebaseerde insig rakende die verskaffer basis.af_ZA
dc.description.versionMastersen_ZA
dc.format.extent98 pagesen_ZA
dc.identifier.urihttp://hdl.handle.net/10019.1/110058
dc.language.isoen_ZAen_ZA
dc.publisherStellenbosch : Stellenbosch Universityen_ZA
dc.rights.holderStellenbosch Universityen_ZA
dc.subjectClusteringen_ZA
dc.subjectSupplier-buyer relationsen_ZA
dc.subjectSupplier segmentationen_ZA
dc.subjectMozambiqueen_ZA
dc.subjectCassava suppliersen_ZA
dc.subjectk-means algorithmen_ZA
dc.subjectAgglomerations, Industrialen_ZA
dc.subjectIndustrial clustersen_ZA
dc.subjectUCTDen_ZA
dc.titleImplementation of clustering techniques for segmentation of Mozambican cassava suppliersen_ZA
dc.typeThesisen_ZA
Files
Original bundle
Now showing 1 - 1 of 1
Loading...
Thumbnail Image
Name:
matshabaphala_clustering_2021.pdf
Size:
6.08 MB
Format:
Adobe Portable Document Format
Description:
License bundle
Now showing 1 - 1 of 1
Loading...
Thumbnail Image
Name:
license.txt
Size:
1.71 KB
Format:
Plain Text
Description: