This project combines U.S. flood insurance claims, population estimates, and climate data (ONI) to build predictive models for flood risk forecasting in Texas. The pipeline includes data preparation, dataset merging, and time series model execution using SARIMAX.
-
Download the NFIP Claims Dataset (CSV) from FEMA.
🔗 https://www.fema.gov/openfema-data-page/fima-nfip-redacted-claims-v2 -
Follow instructions in the
import/README.mdfile to prepare the dataset. -
Place the downloaded file in the
import/directory of this repository.
-
Clone or navigate to the data management repo:
🔗 https://github.com/hirokoclanger/floods/tree/main/code/data-management -
Open and run the following Jupyter Notebooks in order:
1_fetchData.ipynb– loads and pre-processes FEMA and NOAA datasets.2_mergeDataSets.ipynb– aligns claims, population, and ONI datasets into a single time series.
After merging the data, proceed to the modeling phase:
-
Navigate to the
models/directory. -
Run the modeling notebooks:
SARIMAX.ipynb– builds SARIMAX model with ONI as an exogenous regressor.ARIMA.ipynb– builds ARIMA.
- experiements holds the runs and findings
- code holds the files and preprocessing data except the FEMA data since the size exceeds the github limmits