Login

Publications  •  Project Statistics

Glossary  •  Schools  •  Disciplines
People Search: 
   
Title/Abstract Search: 

Dissertation Information for Lixin Yu

NAME:
- Lixin Yu

DEGREE:
- Ph.D.

DISCIPLINE:
- Library and Information Science

SCHOOL:
- SUNY, Albany (USA) (1996)

ADVISORS:
- Ben-Ami Lipetz

COMMITTEE MEMBERS:
- None

MPACT Status: Incomplete - Inspected

Title: Development and evaluation of a framework for assessing the efficiency and accuracy of street address geocoding strategies

Abstract: An important preliminary task in many Geographic Information Systems (GIS) applications is to locate the sites of interest on a map based on their street addresses, a process called geocoding. This process involves comparing input addresses with an established authority file to obtain coordinates. Successful geocoding depends on the characteristics of the input file and the authority file, and the algorithm used in matching the addresses. The algorithm needs to process inexact matches by employing a weighted searching strategy. The weight setting on each address component determines the output quality. No research has examined the relationship between the weight and the output quantitatively, although findings can help GIS users generate better results. In addition, the choice of an authority file has important cost implications. No previous research compares the position precision of the output derived from the two major types of authority files (TIGER-based and zip+4 databases) from the perspective of benefits compared with costs. This research examines these two issues.

The project established and tested a framework for assessing the efficiency and precision of street address geocoding strategies. This framework uses a highly respected (in terms of precision) third party authority file as a reference. Displacement distance between the geocoding results and the true ground locations (ground truth) is used to distinguish mismatches from position errors. A method for categorizing address components and automatically performing analysis using 625 different weighting schemes is introduced.

Three thousand two hundred general addresses were sampled from four New York counties for testing of weighting schemes. Results show that a linear formula can predict the accuracy and matching rate from the weight settings. Weighting schemes with highest matching rates and highest accuracy are identified and listed.

This study found that matching rate and accuracy have a steep reverse linear relationship. In other words, small improvement in matching rate may result in a large decrease in accuracy.

A position precision test produced an interesting finding: the two major types of authority files generate results with limited displacement difference. This result may be used as a reference in selecting authority files for future applications.

MPACT Scores for Lixin Yu

A = 0
C = 3
A+C = 3
T = 0
G = 0
W = 0
TD = 0
TA = 0
calculated 2008-01-31 06:05:37

Advisors and Advisees Graph