Supplier Evaluation Model on SAP ERP Application …

1 Supplier Evaluation Model on SAP ERPA pplication using Machine Learning AlgorithmsManu KohliSchool of Informatics and ComputingIndiana University Bloomington,USAE mail: For business enterprises, Supplier Evaluation is amission critical process. On ERP (Enterprise Resource Planning)applications such as SAP, the Supplier Evaluation process isperformed by configuring a linear score Model , however thisapproach has a limited success. Therefore, author in this paperhas proposed a two-stage Supplier Evaluation Model by inte-grating data from SAP Application and ML algorithms. In thefirst stage, author has applied data extraction algorithm onSAP Application to build a data Model comprising of relevantfeatures.

2 In the second stage, each instance in the data modelis classified, on a rank of 1 to 6, based on the supplierperformance measurements such as on-time, on quality andas promised quantity features. Thereafter, author has appliedvarious machine learning algorithms on training sample withmulti-classification objective to allow algorithm to learn supplierranking classification. Encouraging test results were observedwhen learning algorithms,(DT) and Support Vector Machine(SVM), were tested with more than 98 percent accuracy on testdata sets. The Application of Supplier Evaluation Model proposedin the paper can therefore be generalised to any other otherinformation management system, not only limited to SAP, thatmanages Procure to Pay Terms Machine learning, SAP, ERP, SVM, SupplierEvaluation, Decision Tree, Procurement, Support Vector Ma-chine, Naive Bias, Supplier Ranking,Supervised Learning,HANAI.

3 INTRODUCTIONS upplier Evaluation process has gained importance since lastfew years even though the process is challenging and costly toperform. Setting up decision criteria and Evaluation methodsfor Supplier performance has always been a topic of interest forresearch practitioners [1]. The last two decades have witnesseda growth in the supply chain function wherein the activityof purchase and subsequent Supplier evaluations is stronglycorrelated with the performance of the enterprise. Appropriateselection of suppliers is, therefore, an important prerequisitefor an organization to manage supply chain making with regards to Supplier selection and per-formance Evaluation includes both quantitative and qualitativeassessment factors [2] [3].

4 Several previous types of research have shown the use ofstatistical and mathematical methods for Supplier commonly utilized method is the Data EnvelopmentAnalysis (DEA) that can be used for the measurement ofsupplier efficiency and reduce the total cost of ownership[4]. A study by Hashemi, Karimi, and Tavana [5] suggesteda Model for Supplier optimization using a hybrid approachinvolving a combination of Grey Relational Analysis (GRA)and Analytical Hierarchy Process (AHP).Machine Learning (ML) is used as an alternative techniquethat can be applied to resolve complex classification [6] in a recent study proposed a hybrid Model combiningML and statistical methods to evaluate suppliers.

5 SupportVector Machine (SVM) is one important ML algorithm that isapplied by vast number of researchers to resolve classificationproblems [7]. However, only a handful of studies have beenconducted to date that have evaluated the Application of SVMto perform Supplier Evaluation [8] [9].The majority of small businesses [10] and large enterprisesuse information systems to manage their purchasing 80 percent of fortune 1000 and 60 percent of fortune2000 companies use SAP as their ERP tool [11] to manageprocesses. SAP Application can successfully manage processesand offer controls in functions such as planning of prod-uct, procurement, inventory management, vendor management,customer services and so on [12].

6 SAP Application can alsomanage the procurement process successfully from PROCUREto PAY; however no predictive outcome can be generated fromthe Application that may suggest selection of Supplier based onthe Supplier historical performance has, therefore in this paper proposed a supplierevaluation Model developed on Artificial Intelligence (AI)platform to resolve Supplier selection dilemma. The paper listproblem statement and hypothesis in section 3, data modeland two stage approach to test hypotheses in section 4 andexperiment results in section 5 and 6 of this outcome of the research shows that DT and SVMalgorithms are able to classify Supplier rankings with morethan 97 percent accuracy, allowing Model to be used as adecision support system by procurement department in LITERATUREREVIEWS everal previous studies using decision-making modelssuch as analytic network process (ANP), data envelopmentanalysis (DEA), analytic hierarchy process (AHP), simplemulti-attribute rating technique (SMART) and artificial neuralnetworks (ANN)

7 Have been carried out to perform supplierevaluation and , Xu, Dey and Prasanta [13] carried out literature reviewhighlighting multiple relevant criteria to perform a supplierevaluation. Identification of the best method that could suitsupplier Evaluation is a daunting task, highly dependent onthe industry sector in which an organization of the Supplier Evaluation models mentioned in re-search literature can be classified into four major categoriesnamely the linear weighted models, artificial intelligence (AI)based techniques, mathematical programming models and totalcost the linear weighted Model , multiple criteria attributed torank Supplier can be assigned a certain weight. Supplier perfor-mance is measured as the sum of the values acquired throughmultiplication of the criteria and corresponding weights.

8 Thismodel is simple to implement but mostly depends on thejudgement of an individual to select all the possible criteria andassign weights to them. The linear weighted models includeWeighted point Model or linear weighted mode, analyticalhierarchy process (AHP), and Categorical method. On thecontrary, the total cost models are completely dependent oncosts wherein in addition to the product rate, the indirectcost of an item is also considered. Even in these models, theconcept of subjectivity cannot be models are used for Supplier ranking or selec-tion and are generally complex to implement and have beenpredominantly used to resolve allocation problems. Multi-attribute-decision-making methods include integer program-ming, multi-criteria programming, goal programming, linearprogramming, and mixed integer such as neural networks, quality function deploy-ment, analytic network process, data envelopment analysis(DEA), and fuzzy set theory are also used extensively byvarious researchers [14] to perform Supplier Evaluation .

9 Simicet al. [15] conducted a detailed literature review on fuzzymethods used for Supplier assessment and the present paper, author has used ML and AI techniquesthat have been deployed in various industries to resolveclassification problem but their Application in the Supply ChainManagement (SCM) function is limited. Real world infor-mation systems such as SAP contains Supplier performancemeasurements that can be integrated with AI techniques toclassify and rank suppliers. With the advent of SAP HANAand in-memory databases [16], a large amount of data, bothmaster and transactional, may reside in the real time memoryof SAP Application for analysis and decision making.

10 There-fore, the Application of machine learning algorithms to resolveclassification problems, such as in the case of Supplier ranking,becomes significantly previous research carried out by Lee and Ou-Yang [17]resulted in an accurate predictive Model based on artificialneural networks that could be used for acquiring supportssuch as bid negotiation and making recommendations duringsupplier negotiation process. Wu [6] further presented a hybridmodel encompassing the DEA method for classification ofsuppliers into different clusters based on the efficiency data was used for training the Model , designed combiningdecision tree and neural networks algorithms. The resultingmodel could be used to evaluate new suppliers and displayedencouraging classification et al.