A Study on the Application of Classification and Regression Trees(CART)via an Example of Sweet Potato Data

CCUR > College of Business > Department of Accounting & Graduate Institute of Accounting > periodical articles > Item 987654321/37489

Please use this identifier to cite or link to this item: https://irlib.pccu.edu.tw/handle/987654321/37489

Title:	A Study on the Application of Classification and Regression Trees(CART)via an Example of Sweet Potato Data 分類迴歸樹狀法(CART)在農業應用上之研究--以甘藷資料為例
Authors:	李慕萱
Contributors:	會計系
Keywords:	分類迴歸樹狀法理化性質食用品質特性 CART Eating quality Physico-chemical property Trees classifiers model Misclassification rate
Date:	1996-03
Issue Date:	2017-08-17 13:22:47 (UTC+8)
Abstract:	分類迴歸樹狀法(CART)為一應用在分類問題上，極為簡單且有效之無母數統計方法。在醫學，氣象學、化學、物理學……等專門學科上，皆有廣泛的應用。CART方法之應用，不但能對已收集之資料作精確的估計，且能進一步對未來資料提供預測的功能。 CART方法之基本理論，是將所收集之資料，根據其所有可能之影響變數，作二元樹狀之分類。其方法為先找出一個分類點變數，以此將資料分為二大類，再各別對此二類資料，找出其各別之分類點，……，如此反覆執行，直到資料無法再細分或可歸類為某一等級為止。據此，則可建立一個二元樹狀結構模式。根據分類點之變數，可將資料作一適當之分類。本研究之主要目的為介紹CART方法之基本理論，並將此方法應用在農業資料上，以期能對農作物之食用品質之分類問題，提供一簡單且有效之分析方法。茲以甘藷食味和理化性質資料為例，應用CART方法，根據甘藷之理化性質變數，以建立一個二元樹狀結構模式，用以推論其食用品質特性之等級分類。根據此甘藷資料所建立之樹狀分類模型，對甘藷食用品質特性所作之分類，其錯誤估計比率為0.125。 The application of CART in classification problems is very easy and flexible. It has been widely used in medical science, chemistry, meteorology, physics, ……, etc. In agricultural application, CART is useful to classify the products to different classes of quality depending on a variety of variables. The purpose of this paper is to introduce the use of CART in agricultural applications. First we will introduce the development of CART briefly. Secondly, the method of CART in agricultural application is presented. Finally, an example of the sweet potato data is given for illustration. The trees classifiers model is constructed for the sweet potato data so it can be used to classify the different scales of taste score of eating quality based on the physicochemical property variables. This model provides a very easy way to the selection of sweet potatoes in different scales of taste score of eating quality depending on the situations of a few variables. The size of the data limits the use of more variables and splits in the trees model. The estimate of misclassification rate for this model is low at 0.125.
Relation:	中華農學會報 173 民85.03 頁60-68
Appears in Collections:	[Department of Accounting & Graduate Institute of Accounting] periodical articles

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	138	View/Open

View Licence

Loading...