The importance of the project selection phase in any six sigma initiative cannot be emphasized enough. The successfulness of the six sigma initiative is affected by successful project selection. Recently, Data Envelop...The importance of the project selection phase in any six sigma initiative cannot be emphasized enough. The successfulness of the six sigma initiative is affected by successful project selection. Recently, Data Envelopment Analysis (DEA) has been proposed as a six sigma project selection tool. However, there exist a number of different DEA formulations which may affect the selection process and the wining project being selected. This work initially applies nine different DEA formulations to several case studies and concludes that different DEA formulations select different wining projects. Also in this work, a Multi-DEA Unified Scoring Framework is proposed to overcome this problem. This framework is applied to several case studies and proved to successfully select the six sigma project with the best performance. The framework is also successful in filtering out some of the projects that have “selective” excellent performance, i.e. projects with excellent performance in some of the DEA formulations and worse performance in others. It is also successful in selecting stable projects;these are projects that perform well in the majority of the DEA formulations, even if it has not been selected as a wining project by any of the DEA formulations.展开更多
通过对Spark并行计算框架以及投影树算法的分析,提出基于Spark的投影树频繁项集挖掘算法(projection tree algorithm based on Spark,PTBS)。为解决重复遍历数据集带来的损耗,改变数据的存储结构;利用广播变量对频繁二项集的产生进行改...通过对Spark并行计算框架以及投影树算法的分析,提出基于Spark的投影树频繁项集挖掘算法(projection tree algorithm based on Spark,PTBS)。为解决重复遍历数据集带来的损耗,改变数据的存储结构;利用广播变量对频繁二项集的产生进行改进,降低运算量,减少集群节点间通信的消耗;将Apriori先验性质添加到候选项集的筛选过程中,减少节点间的通信量。将算法与Spark平台所具备的特殊性能相结合,得到基于Spark的投影树关联规则挖掘算法,实验结果表明,该算法具备良好的可扩展性,适宜于处理大规模数据。展开更多
文摘The importance of the project selection phase in any six sigma initiative cannot be emphasized enough. The successfulness of the six sigma initiative is affected by successful project selection. Recently, Data Envelopment Analysis (DEA) has been proposed as a six sigma project selection tool. However, there exist a number of different DEA formulations which may affect the selection process and the wining project being selected. This work initially applies nine different DEA formulations to several case studies and concludes that different DEA formulations select different wining projects. Also in this work, a Multi-DEA Unified Scoring Framework is proposed to overcome this problem. This framework is applied to several case studies and proved to successfully select the six sigma project with the best performance. The framework is also successful in filtering out some of the projects that have “selective” excellent performance, i.e. projects with excellent performance in some of the DEA formulations and worse performance in others. It is also successful in selecting stable projects;these are projects that perform well in the majority of the DEA formulations, even if it has not been selected as a wining project by any of the DEA formulations.
文摘通过对Spark并行计算框架以及投影树算法的分析,提出基于Spark的投影树频繁项集挖掘算法(projection tree algorithm based on Spark,PTBS)。为解决重复遍历数据集带来的损耗,改变数据的存储结构;利用广播变量对频繁二项集的产生进行改进,降低运算量,减少集群节点间通信的消耗;将Apriori先验性质添加到候选项集的筛选过程中,减少节点间的通信量。将算法与Spark平台所具备的特殊性能相结合,得到基于Spark的投影树关联规则挖掘算法,实验结果表明,该算法具备良好的可扩展性,适宜于处理大规模数据。
