| Title |
Model for predicting dog stay time at “Penkta Koja” animal shelter / |
| Authors |
Litvaitytė, Ignė ; Šiugždaitė, Simona |
| DOI |
10.15388/VGISC.2024.II |
| Full Text |
|
| Is Part of |
19th Prof. Vladas Gronskas international scientific conference, 29th of November, 2024, Kaunas, Lithuania : abstract book.. Vilnius : Vilniaus universiteto leidykla. 2024, p. 39-40 |
| Keywords [eng] |
dog shelter ; machine learning ; XGBoost regression ; adoption prediction ; data analysis |
| Abstract [eng] |
The “Penkta Koja” dog shelter faces challenges with prolonged stays for some dogs awaiting adoption. This study aims to develop a machine learning model to predict a dog’s shelter stay duration based on available attributes such as gender, age, size, color, and arrival/departure quarters. Utilizing the XGBoost regression algorithm, the model processes categorical variables through OneHotEncoding and standardized numerical features. With hyperparameter tuning via GridSearchCV, the best configuration achieved an average absolute error (MAE) of 59.96 days and an R² of 0.18, indicating low prediction accuracy and room for improvement. For example, the model predicted a 99.11-day stay for a medium-sized, 2-month-old black dog. While initial results provide insights, integrating additional features like temperament or health status could enhance accuracy, offering a practical tool for the shelter to optimize care and adoption efforts. |
| Published |
Vilnius : Vilniaus universiteto leidykla |
| Type |
Conference paper |
| Language |
English |
| Publication date |
2024 |
| CC license |
|
