Classification metrics for pet adoption prediction with machine learning

Abstract

Millions of pets are temporarily placed in shelters, making it challenging for shelters to ensure pets find permanent homes. High adoption rates are crucial for animal welfare and the sustainability of shelter operations. This study aims to identify key factors influencing pet adoption and create classification metrics using five machine learning (ML) classification model approaches to predict the likelihood of pet adoption, to find the best model performance for each analysis. The dataset was obtained from several features related to animal characteristics and adoption conditions. The results of the study present classification of metric models that indicate decision tree and random forest (RF) as the most effective models with superior performance in terms of accuracy and class separation ability. Further research provides initial exploration of ML models that are not only limited to classification models but also model integration into internet of things (IoT) systems for the implementation of a pet adoption prediction system based on ML inference. The implementation of ML classification models helps improve the efficiency of animal adoption programs and optimize shelter operations, ultimately increasing the chances of successful pet adoption. The results of the study provide insights into factors influencing pet adoption, minimizing the length of stay (LOS) in shelters, and contribute to practitioners/ researchers as a reference for exploring new related factors and exploring the performance of ML models, especially classification models.