應用集成式學習法於時間序列預測之研究

文化大學機構典藏 CCUR > 商學院 > 資訊管理學系暨資訊管理研究所 > 博碩士論文 > Item 987654321/33732

jsp.display-item.identifier=請使用永久網址來引用或連結此文件: https://irlib.pccu.edu.tw/handle/987654321/33732

题名:	應用集成式學習法於時間序列預測之研究 A Study of Ensemble Learning for Time Series Forecasting
作者:	洪國展
贡献者:	資訊管理學系
关键词:	時間序列預測分析集成式學習堆疊泛化集成 time series forecast analysis ensemble learning stacking
日期:	2016
上传时间:	2016-08-18 09:39:10 (UTC+8)
摘要:	在現今科技發達的環境中，網際網路普及與交通運輸發達，人與人之間的聯繫已經變得越來越緊密，國與國之間的交流變得沒有界線，全球化成為一種現象，此外經濟全球化最為顯著。在經濟全球化的快速發展，整個經濟環境的改變與市場的競爭下，企業面臨的競爭與壓力有別於以往，企業必須順應時代改變，有效掌握與瞭解未來趨勢，才是致勝的關鍵。為了瞭解未來趨勢變化，預測分析備受各企業組織與學術界所重視，其中以時間序列預測分析最為熱門。因此本研究利用AREMOS台灣經濟統計資料庫所提供之工業產品產銷存量資料作為分析來源，透過資料探勘的技術建立集成式學習預測模型，首先建立類神經網路、支援向量機、隨機森林之單一分類器，接著以線性回歸與類神經網路演算法作為集成分類器，建立一個預測能力良好的集成式學習預測模型。實驗結果顯示，集成式學習預測模型整體的表現大多優於單一預測模型，其中線性回歸集成式學習預測模型預測能力最為優異。 In this modern age of technologically advanced environment, the popularization of Internet, and accessible transportation become a connection between people and become closer. Exchanges between countries become instant and seemingly borderless. Globalization has become a phenomena in which economic expansion is more visible. Economic globalization, the rapid development of the whole economic environment has changed the market competition. Enterprises are facing tournament and pressure that are different from the previous age. Now they must confirm the changes of effective control and understand future trends. This is crucial for their success. Forecast analysis is an important method for enterprises and academic to understand the future trends. One of the most popular algorithms is Time Series Forecast Analysis. In this study, the Industrial products sales data provided by AREMOS are used as the data of analysis. We present an ensemble data mining approach for forecast analysis. First, Individual classifiers are built based on three data mining techniques: Neural Networks, Support Vector Machines and Random Forest. Their outputs are then combined by two stacking methods: Linear Regression and Neural Networks. Experiment results show that the ensemble model outperforms the individual classifier models and the linear regression ensemble model performs the best overall.
显示于类别:	[資訊管理學系暨資訊管理研究所 ] 博碩士論文

文件中的档案:

档案	描述	大小	格式	浏览次数
index.html		0Kb	HTML	291	检视/开启

检视Licence

在CCUR中所有的数据项都受到原著作权保护.

数据加载中.....