DEA-SBM 모델과 EM 군집 분석에 근거한 자동차 공급사슬 효율성 평가 및 벤치마킹 방안

Abstract

The objective of this paper is to present the model for evaluating the supply chain efficiency and benchmarking process of the component suppliers in Korean automotive industry. The proposed method is theoretically based on the combination of the slacks based super efficiency model (DEA-SBM) and the expectation maximization (EM) algorithm clustering analysis. DEA-SBM is one of many data envelopment analysis (DEA) with input and output criteria to evaluate the relative efficiency in order to rank the decision making unit (DMU). EM algorithm makes cluster based on probability considering the maximization of log-likelihood. This method can be differentiated from previous approaches in the literature. This study uses time series data from 2012 to 2014 of 64 tier-one Korean domestic automotive chassis parts suppliers and adopts the DEA-SBM model in order to increase the analysis's accuracy and EM clustering to classify the suppliers by the actual input resources. It also suggests the practical stepwise benchmarking paths to the inefficient firms. In this study, three input factors are selected: number of employees, assets and cost of goods sold (COGS). The output factors include two financial and three non-financial data: sales, operating profit and three kinds of customer’s evaluation scores about suppliers, i.e quality, delivery and technology. Firstly, we conduct the EM clustering to classify the suppliers depending on three input factors, resulting in seven clusters. We analyze the cluster changes of each supplier over time so as to monitor the cause of changes. Then, we calculate the super-efficiency scores of the suppliers using DEA-Solver Pro 9.0 software. We also conduct the analysis about the changes of efficiency scores over time. From this analysis, we apparently recognize which suppliers get better or worse over three years. For the sake of the comprehensive evaluation, we use a hybrid approach, i.e. the efficiency analysis based on the EM cluster as well as on the industrial sectors. Since the value chain is totally different depending on the industrial sectors. The industrial sector is also a key perspective like EM cluster. The best efficient EM cluster was cluster 7 and the worst was cluster 4. On the other hand, the best efficient industrial sector was assembly sector and the worst was fastener sector. From the results of the analysis, we can find out the most inefficient cluster and industrial sectors which need to improve efficiency to increase the overall supply chain performance. In the case study, we selected the fastener sector as the improvement targets. We suggested the stepwise improvement paths. One stage is the efficiency improvement within the same EM cluster, which anticipated fast benchmarking and improvement due to the similarity of the firms' scales. Other stage is the improvement between different EM clusters. To obtain the final improvement targets, we adopt the DEA projection analysis within the selected industrial sector. We illustrate the benchmarking paths as the case of DMU #7 belonging to the fastener sector and EM cluster 5. As we have seen in the case study, this study provides the inefficient firms with a stepwise benchmark approach considering industrial sectors with final improvement target in terms of input and output factors to reach the highest level of efficiency frontier in the same industrial sector. With this proposed method, the buyer, specifically a car manufacturer, can precisely evaluate their suppliers, strategically select the improvement targets, practically determine improving direction and suggest effective benchmarking paths to suppliers.