A Decision-Tree Based Recommending Approach for Enterprise Cloud Service

• Main Authors: Zih-Neng Lin, 林子能
• Other Authors: 張瑋倫
• Format: Others
• Language: zh-TW
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/67964732167976326104

碩士 === 淡江大學 === 企業管理學系碩士班 === 104 === According to the prediction of XO Communications, there will be 86% of businesses building hybrid cloud in 2017. Both in technology and practice, The proportion that enterprises use hybrid cloud for deployment has been significantly improved in terms of technical and practical perspectives. In addition, hybrid cloud is one of the deployment ways enterprises prefer to use currently since it has the benefits of public cloud and private cloud simultaneously. That is, most enterprises utilize their cloud services by hybrid cloud. Recently, economic depression makes enterprises more concern about the low cost and flexibility of cloud applications. More and more companies provide cloud computing services. This research proposed two research goals in order to take into account the high failure rate of selecting inadequate cloud projects as well as diverse and complex factor selection. First, we identify key factors of selecting cloud services through literature and experts. Besides, we use machine learning technique (decision tree) to establish an effective decision model. In this study, we surveyed literature to identify key factors and inquired experts to validate the designed questionnaire. The questionnaires for business users are collected for C4.5 algorithm to obtain a decision model. In the study, experts confirmed content validity of selected 17 key factors through literature. In addition, 35 rules were established in the decision tree and result showed the common cloud products used in Taiwan were Amazon AWS EC2 and IBM Smart Cloud. In particular, IBM Smart Cloud was used in education business mostly. Furthermore, we discovered the key factors for selecting different cloud products. For example, there are 28 with the properties of information security capabilities. Appropriate cloud products can be recommended according to different needs. The accuracy of our model is 85% and the recall rate is 86% of the decision tree, which achieves the high standard of performance. Compared to the past researches with similar methods, the prediction of the decision tree model is effective for cloud solution selection in practice.