請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/72865
標題: | 單邊界時間序列預測 Unilateral Boundary Time Series Forecasting |
作者: | Chao-Min Chang 張櫂閔 |
指導教授: | 林守德(Shou-De Lin) |
關鍵字: | 時間序列,預測,機器學習,低估,高估,雜訊, Time Series,Forecasting,Machine Learning,Underestimation,Overestimation,Noise, |
出版年 : | 2019 |
學位: | 碩士 |
摘要: | 在許多的現實應用問題中,都出現了單邊界時間序列問題。例如,空汙檢測系統、交通預測系統、物聯網感測、以及潛在客戶預測問題。在這篇論文中,我們嘗試解決這個問題。我們提出一個完整的機器學習模型以解決單邊界時間序列問題。我們針對單邊界時間序列中多個不同的面向,一一提出了解決方法。接著,我們模擬了現實應用問題中多種低估的情形,並且我們實驗多個不同的資料,演示我們提出的方法的優勢。 There are many single boundary time series forecasting problems in the real-world applications. For example, air quality censors, traffic estimation, IoT censors, potential customer problem, etc. In this paper, we tried to solve this problem. We proposed a generalized model for unilateral boundary time series forecasting. Three different methods are proposed and combined to deal with each aspects of the single boundary time series forecasting. Besides, we simulated several underestimation situation with real-world application examples and conducted experiments on three different data sets to demonstrate the general performance of our proposed model. In addition, we conducted experiments on both tree-based model and deep learning model to show the generality of the models that our methods can be implemented on. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/72865 |
DOI: | 10.6342/NTU201901686 |
全文授權: | 有償授權 |
顯示於系所單位: | 資訊工程學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-108-1.pdf 目前未授權公開取用 | 908.18 kB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。