Determining Pets' Length of Stay (LOS) in a Shealter: A Data Mining Approach

• Main Author: Siphamandla Musa Dlamini
• Other Authors: Dr. Yo-Ping Huang
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/pw7v65

碩士 === 國立臺北科技大學 === 電資學院外國學生專班 === 103 === The continuous advancement of technology and health care services continues to ensure an increased life expectancy in both humans and animals. As in numerous areas of life such as medical domain, business and as well as in shelter management, the volumes and complexity of data collected and stored continue to grow every day. While most pet owners are clear about the immediate joys that come with sharing their lives with companion animals, many remain unaware of the physical and mental health benefits that can also accompany the pleasure of playing with or snuggling up to a furry friend. It is only recently that studies have begun to scientifically explore the benefits of the human-animal bond. The goal of this empirical study is to examine pet adoption records and determine whether certain variations in physical traits such as breed, color, size, age, weight, etc. are related to greater adoption success. Based on a comprehensive analysis of such data, it may be possible to come up with new hypotheses and find statistical evidence for existing hypotheses, in particularly around the prediction of Length of Stay (LOS) and the physical attributes of pets. In this thesis, we use data mining classification techniques such as decision trees and association rules in order to investigate influential pets’ traits that may contribute to a stay of a pet in the shelter house. The data mining methods are used to build models for finding frequent and interesting patterns in data. In pet shelter management these would be very applicable for building accurate classifier models which predicts the LOS and may be key to the decision makers in the shelter management value chain. In this study we used a four year annual pet data collected from the animal shelter (Society for Prevention of Cruelty to Animals) in Tampa Bay, Florida, USA. Each year over 4,000 records of pet information are collected. The results will help local shelters better allocate limited sources, reduce financial implications and formulate strategies to increase pet adoption, assist potential adopters in better decision making to find an ideal pet and also assist government in understanding pet adoption trends for better policy formulation that safeguard pets and shelter management.