- 學位論文
數據導向研究:物料使用狀態之大數據與預測分析
A data-driven research: Big data and predictive analytics for material usage status
摘要
近年來數據科學有著極大的發展，其中包括數學，統計，電腦科學，與行為科學等多項跨領域的技術突破。許多供應鏈的專家積極地蒐集資料與數據，從客戶的訂單採購，生產製造與存貨，至最終出貨，期望利用各種新興科技與技術，如工業4.0，以精準掌握顧客行為，提升預測的準確率。為了達成上述的需求，企業積極進行數位化轉型，將訂單數據化，建構與挖掘數據並進行分析預測，但往往無法掌握數據之品質。換句話說，面對海量的資料，目前雖有許多完整的統計分析工具為專業人員提供了分析數據的動力，但基於數據的建構及處理仍是大數據技術最重要的環節。 由於分析工具中，數據品質將影響結果甚鉅，因此有效的了解所蒐集的數據品質與解決的問題有高度的影響。一般而言，數據品質之定義特性相當廣泛，本研究之目的為找出物料管理資料所需之數據品質，基於其特性之資料進行建模預測，其預測結果是否會因其中一項特性的缺陷導致需求預測的結果有大幅的影響。研究中也將利用案例評估數據品質對於需求預測之影響，並以改善模型之方式，彌補數據品質之缺陷，期望本研究的成果能夠對業界在大數據分析中有所助益。
並列摘要
In recent years, data science has developed tremendously, including multiple technological breakthroughs in mathematics, statistics, computer science, and behavioral science. Many supply chain experts actively collect information and data, from customer order purchase, manufacturing and inventory, to final shipment, hoping to use various emerging technologies and technologies, such as Industry 4.0, to accurately grasp customer behavior and improve forecast accuracy. In order to achieve the above-mentioned needs, companies are actively carrying out digital transformation, digitizing orders, constructing and mining data, and analyzing and predicting data, but they often fail to grasp the quality of the data. In other words, in the face of massive data, although there are many complete statistical analysis tools that provide professionals with the power to analyze data, data-based construction and processing are still the most important aspects of big data technology. Since the data quality in the analysis tools will greatly affect the results, an effective understanding of the quality of the collected data and the problem to be solved has a high impact. Generally, the defining characteristics of data quality are quite extensive. The purpose of this study is to find out the data quality required for material usage data, and to model and predict based on the data of its characteristics. Will the predicted result be due to a defect in one of the characteristics? The result of demand forecasting has a substantial impact. The research will also use cases to evaluate the impact of data quality on demand forecasting, and improve the model to make up for the defects of data quality. It is hope that the results of this research will be helpful to the industry in big data analysis.