結合領域知識與機器運算之新的特徵選取方法: 應用於財務危機預警預測之問題 ; Novel feature selection methods to Financial Distressed Prediction problem

NCUIR > College of Electrical Engineering & Computer Science > Software Engineer > Electronic Thesis & Dissertation > Item 987654321/48302

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/48302

Title:	結合領域知識與機器運算之新的特徵選取方法: 應用於財務危機預警預測之問題;Novel feature selection methods to Financial Distressed Prediction problem
Authors:	張懷倫;Huai-lun Chang
Contributors:	軟體工程研究所
Keywords:	遺傳演算法;財務危機預測;特徵選取;genetic algorithm;wrapper method;Financial Distressed Prediction;Feature Selection
Date:	2011-07-21
Issue Date:	2012-01-05 14:49:54 (UTC+8)
Abstract:	在目前眾多的研究議題中，特徵選取(Variable and feature selection)已經是一個越來越令人關注的議題。尤其是當我們收集樣本的特徵集(Feature sets)成千上百的增加的時候，一個好的特徵選取方法可以使得結果令人滿意。本論文提出了一個概念，此概念是嘗詴去結合專家意見(Expert recommendation)與機器學習演算法(Machine Learning Algorithm)後，創造出一種混合型的特徵選取方法(Novel feature selection methods)，並且使用預測財務危機公司(Financial Distressed Prediction，簡稱FDP)此問題當作案例做為實驗證實。本論文的貢獻在於對於特徵選取這個議題而言，我們提供了兩個新的方法：Advanced wrapper method & Mix of Expert and Machine(MEM)。而這兩個方法對於應用在非結構化的商業問題上(unstructured nature of the business problems)有著比貣以往的方法更佳的結果-擁有更勝於以往的預測準確率以及為數少量的推薦特徵集。 Variable and feature selection is an important issue in plenty of issues, especially feature sets is growing up violently. A good variable and feature selection will have bearing on performance of result. In this paper, we apply a new concept that combines expert recommendation and machine learning algorithm to create a novel feature selection, and utilize the financial distress prediction problem as a study case to prove our idea. We apply two methods that Advanced wrapper method & mixed of expert and machine (MEM) to applicate in nonstructed business problem and believe this proposed methods be better performance than original methods included predictor accuracy and few feature set.
Appears in Collections:	[Software Engineer] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	1113	View/Open

社群 sharing

Loading...