Forecast Shelter Adoption and Euthanasia Rate Trends

ISEF Category: Animal Sciences

Subcategory: Other  ·  Difficulty: Advanced  ·  Setup: Home Setup  ·  Time: 1 to 2 Months

The Hook

A rainy month can change how many animals enter a shelter, and how many leave. That means adoption and euthanasia rates are not random. You can test whether weather and local economic conditions help predict those swings. This turns a real shelter problem into a data project you can measure.

What Is It?

This project builds a machine-learning model, which is a program that learns patterns from past data and uses them to make predictions. In this case, the model looks at shelter intake records, local weather, and economic indicators, then tries to forecast adoption and euthanasia rates. Think of it like a recipe made from several ingredients. Shelter records tell you what happened inside the shelter, while weather and economic data tell you what was happening outside it.

The key idea is that shelter outcomes may change with the season, the job market, or local stress on families. For example, a shelter may see more intake during certain weather patterns or fewer adoptions during tough economic periods. Your job is to test whether those outside signals actually improve predictions. If they do, you have shown that shelter outcomes connect to broader community patterns, not just the shelter itself.

Why This Is a Good Topic

This is a strong science fair topic because you can test it with public or easily requested data, and the answer is not obvious before you analyze it. You can compare simple models against stronger ones, which gives you real evidence instead of a guess. The project also connects to animal welfare, public policy, and community health, so the work feels useful beyond the fair. A student can learn data cleaning, feature selection, model evaluation, and how to explain results clearly.

Research Questions

• How does monthly rainfall affect the adoption rate in shelters in one county?
• What is the effect of local unemployment rate on euthanasia rate after controlling for intake volume?
• Does adding weather data improve next-month prediction accuracy compared with shelter records alone?
• To what extent do seasonal temperature swings explain changes in adoption-vs-euthanasia outcomes?
• Which feature set, intake only, weather only, or economic only, gives the lowest prediction error?
• How does shelter size change model accuracy across different cities or counties?

Basic Materials

• Public animal shelter intake and outcome records.
• Local weather data from a government source or open data portal.
• County-level economic indicators such as unemployment, income, or housing stress.
• Laptop with internet access.
• Spreadsheet software for cleaning and sorting records.
• Python notebook or Google Colab for analysis.
• Data dictionary or codebook for each dataset.
• Charting tool for line graphs and bar charts.

Advanced Materials

• Historical shelter database with intake, adoption, transfer, and euthanasia fields.
• Longer time series of weather and economic indicators matched to shelter records.
• Python with pandas, scikit-learn, and statsmodels.
• R with tidymodels or caret for model comparison.
• GIS software for mapping shelters and regions.
• Secure storage for sensitive records.
• Version control repository for tracking code changes.
• Institutional review guidance if any protected data is included.

Software & Tools

• Python: Cleans records, builds features, and trains prediction models.
• Pandas: Joins shelter, weather, and economic tables into one analysis file.
• Jupyter Notebook: Keeps code, charts, and notes together while you test ideas.
• scikit-learn: Trains baseline models and compares prediction performance.
• Google Colab: Runs Python in a browser if your laptop is slow.

Experiment Steps

1. Define the exact outcome you want to predict, such as adoption rate, euthanasia rate, or next-month changes.
2. Choose one time scale and one geographic scale so every dataset lines up the same way.
3. Build a simple baseline model using shelter intake data alone, then test whether weather and economic features improve it.
4. Plan controls for missing records, seasonal patterns, and differences between shelters so your comparison stays fair.
5. Decide how you will judge success, using cross-validation, error metrics, and a feature-importance check.

Common Pitfalls

• Mixing daily weather data with monthly shelter outcomes, which blurs the time scale and weakens the model.
• Using adoption and euthanasia labels that mean different things across shelters, which makes the outputs hard to compare.
• Letting the model train on future months by accident, which makes accuracy look better than it really is.
• Forgetting to normalize for shelter size, which lets large shelters dominate the patterns.
• Treating correlation as causation, which can turn a prediction result into an unsupported claim about why outcomes changed.

What Makes This Competitive

A class-level version stops at one model and one score. A stronger project compares several algorithms, holds out an entire shelter or year, and checks whether the pattern still holds across different regions. You can also test whether weather and economic data add value beyond intake counts alone. That kind of careful validation and comparison makes the work much stronger.

Project Variations

• Predict intake spikes instead of adoption rates using the same weather and economic features.
• Compare urban and rural shelters to see whether the strongest predictors change by region.
• Replace census-style economic indicators with housing stress or rent data to test whether local pressure improves prediction.

Learn More

• PubMed: Search review articles on animal shelter outcomes, adoption patterns, and humane euthanasia.
• NOAA Climate Data Online: Download local weather and seasonal data for your model.
• U.S. Census Bureau American Community Survey: Find county-level income, poverty, and housing indicators.
• USDA Economic Research Service: Look up regional economic context and rural-urban measures.
• Your city or county open data portal: Search for animal shelter intake and outcome records if the shelter publishes them.