This project was developed as part of the final Capstone for the HarvardX Data Science Professional Certificate.
It explores the prediction of shelter animal outcomes (e.g. Adoption, Transfer, Euthanasia)
at the Austin Animal Center, based on intake-related features.
The objective is to predict shelter animal outcomes based on intake data such as
animal type, condition, and intake type.
Model performance is evaluated using accuracy, confusion matrix, and feature importance.
All steps follow the edX Honor Code.
The dataset is provided by the Austin Animal Center,
hosted on Kaggle. It contains detailed records of animal intakes and outcomes.
Please download it manually from Kaggle and use it locally for testing and report generation.
👉 https://www.kaggle.com/datasets/aaronschlegel/austin-animal-center-shelter-intakes-and-outcomes
| File | Description |
|---|---|
01_load_data.R |
Load and inspect the shelter dataset |
02_explore_data.R |
Exploratory data analysis (EDA): distributions, NA overview |
03_model_baseline.R |
Baseline model: predict most frequent outcome ("Adoption") |
04_model_randomforest.R |
Random Forest model with 5-fold CV and feature importance |
05_compare_models.R |
Comparison of baseline vs. Random Forest (accuracy & plots) |
06_final_model.R |
Final model application without CV, final evaluation |
07_final_pipeline.R |
Complete pipeline with all steps and explanatory comments |
chooseyourproject_report.Rmd |
Final R Markdown report (edX-compliant) |
chooseyourproject_report.pdf |
Rendered PDF version for submission |
chooseyourproject_report.html |
Rendered HTML version |
LICENSE |
MIT License for reuse |
.gitignore |
Excludes data files and system folders |
README.md |
This project overview |
| Metric | Value |
|---|---|
| Baseline Accuracy | 42.18 % |
| Random Forest Accuracy | 58.09 % |
| Absolute Improvement | +15.91 pp |
| Relative Improvement | +37.7 % |
| Final Model Trees | 500 |
Random Forest classifier clearly outperformed the naive baseline.
Most important predictors: intake_type, sex_upon_intake, intake_condition.
- R 4.x or newer
- RStudio
- Required R packages:
tidyversecaretrandomForestscalestidytext(for variable importance visualization)
The final report submitted for edX is available in two formats:
It includes all modeling steps, evaluations, plots, and interpretations.
This project was created by Yvonne Kirschler
and is licensed under the MIT License.
If you reuse code from this repository, please provide proper attribution.
GitHub profile: @alunera-data
LinkedIn: Yvonne Kirschler
This project was developed independently.
ChatGPT (OpenAI) was used to support structure, planning and phrasing.
All modeling, evaluation and reporting were performed and reviewed by the author.