LOGISTIC REGRESSION FOR CLASS~F~CATION OF TEXT DOCUMENTS

دانشگاه آزاد اسلامی اصفهان(خوراسگان)

منو

رکورد قبلی

رکورد بعدی

" LOGISTIC REGRESSION FOR CLASS~F~CATION OF TEXT DOCUMENTS "

Document Type	:	Latin Dissertation
Language of Document	:	English
Record Number	:	151446
Doc. No	:	ET23238
Main Entry	:	JACK R.BRZEZINSKI
Title Proper	:	LOGISTIC REGRESSION FOR CLASS~F~CATION OF TEXT DOCUMENTS
Note	:	This document is digital این مدرک بصورت الکترونیکی می باشد
Abstract	:	The volume of the information availabIe in the digital fonn continues to increase.Currently there is about 10" bits of information available on the Internet Thereis an increasing need for tools capable of managing large data sets. Mormationcan be stored in many forms. This thesis concentrates on text based informationand examines problems fundamental for one of the Information Retrievd tasks.Automatic text classification, the assignment of text documents to two or morepredefined categories, is an important and difficult problem in InformationRetrieval.Wis research focuses on a machine learning algorithm for text classification.Logistic regression for binary classification and polytomous logistic regressionfor mdtiple ~Iassification belong to the group of inductive learning algorithms.We concentrate on two basic issues. The first issue involves development oflogistic regression models and polytomous logistic regression models. Wedevelop heuristics allowing for bunding very accurate classifiers. Secondly, weaddress computational cost of modeling of high dimensional data We developheuristics, which dlow to reduce the computational cost of model developmentwithout a significant reduction of the accuracy.Model evaluation is an important issue for Statistics and Machine Learningcommunities. In order to provide a fair comparison between different modelingalg~rithms we apply meta cross-validation to compare....-....,....-...,..tested for theQ1 PC1 bus cardBoth these projects mere sofixare des elopment efforts tonards contributing to dlfferentaspects of Roboucs and lZ1echatronics projects m the Controls and Roboucs Group..
Subject	:	Electericl tess
	:	برق
electronic file name	:	TL46474.pdf
Title and statement of responsibility and	:	LOGISTIC REGRESSION FOR CLASS~F~CATION OF TEXT DOCUMENTS [Thesis]