臺灣博碩士論文加值系統

目前雖已有許多用於機器學習的工具，但多不適合用於工業資料分析。分析工業資料常需要處理大量資料的問題，且由於有時難以在分析過程中看出特徵值與目標之間的關聯，常需要嘗試各種模型跟預處理方式的組合才能找出理想的模型。
為了節省分析工業資料時花在熟悉工具以及嘗試不同資料處理方式與模型上的時間，我們提出一個用於處理工業資料的平台，由於其專為工業資料分析設計，平台的功能較一般通用的機器學習工具簡單，無須花大量時間學習如何使用，使用者可以使用此平台提供的多種常用於工業資料的處理方法，透過腳本快速撰寫各分析步驟希望嘗試的資料處理方法，交由平台自動組合各步驟，最後挑出最適合的模型，並自動生成部署用的腳本與方法權重。

Nowadays, there are many tools that can be used on machine learning tasks. However, most of them are not good at industrial data analysis tasks. When analyzing industrial data, we usually have to process big data problem and sometimes not easy to find relation between feature and target data during analysis procedure. Thus, we have to try lots of combinations of data processing methods and training models to find an ideal model.
To avoid consuming time on learning how to use new machine learning tools and trying various combination of data processing methods and training models. We provide a platform for doing industrial data analysis. Because of it is designed to do industrial data analysis, the functionality of it is simpler than general-purposed machine learning tools and, thus, cost less time to learning how to use it. Users can write scripts to describe methods they want to try on every step easily using data processing methods provided by the platform, which are usually used to process industrial data. The platform would combine all the steps with different methods automatically and find the best model then provide you scripts and weights of functions for deployment.

中文摘要 i
英文摘要 ii
誌謝 iii
目錄 iv
表目錄 vi
圖目錄 vii
一、 介紹 1
二、 相關研究 2
三、 架構 3
3.1 輸入腳本文件 4
3.1.1 Data檔（*.data） 4
3.1.2 Module檔（*.module） 4
3.1.3 Script檔（*.script） 5
3.1.4 Config檔（*.cfg） 6
3.2 平台架構 6
3.2.1 Script 6
3.2.2 ScriptCell 7
3.2.3 Module 7
3.2.4 Library 8
3.2.5 DataLoader 8
3.2.6 ResourceManager 8
3.3 初始資料讀取 9
3.4 自訂方法 10
四、 實驗結果 10
4.1 資料集1 － 濺鍍製程 11
4.2 資料集2 － 生產線 14
五、 總結 17
參考文獻 18
附錄一 19

[1]B.Lin,B.Recke,J.K.H.Knudsen,S.B.Jørgensen,Asystematicapproachforsoftsensor development. Computers and Chemical Engineering, 31, 419–425, 2007.
[2]Janez Demšar, Tomaž Curk, Aleš Erjavec, Črt Gorup, Tomaž Hočevar, Mitar Milutinovič, Martin Možina, Matija Polajnar, Marko Toplak, Anže Starič, Miha Štajdohar, Lan Umek, Lan Žagar, Jure Žbontar, Marinka Žitnik, Blaž Zupan,“Orange: Data mining toolbox in python,” Journal of Machine Learning Research, vol. 14, pp. 2349–2353, 2013.
[3]M. Toplak, G. Birarda, S. Read, C. Sandt, S. M. Rosendahl, L. Vaccari, J. Demšar, and F. Borondics, “Infrared orange: Connecting hyperspectral data with machine learning,” Synchrotron Radiation News, vol. 30, no. 4, pp. 40–45, 2017.
[4]M. S. d. R. Luca Rebuffi, “Oasys (orange synchrotron suite): an open-source graphical environment for x-ray virtual experiments,” 2017.
[5]RapidMiner, Inc. RapidMiner. Retrieved June 12, 2019, from https://rapidminer.com/