By Pujika Kumar, Ruining Feng, Cameron Lewis, Emily Park, Sri Shatagopam, Yue Wu, Siyuan Zhang
We will be investigating the consequences of natural disasters both on a regional and global scale in terms of human and economic impact (i.e. from deaths and homeless statistics to economic damage) based on the geographical origin, timeframe of the disaster, etc.
As a non-profit organization we want our end users/donators to understand the importance of our work. Our goal is to convince English speaking potential donors to donate to our company so we can continue our company’s mission: “help others through providing humanitarian aid.” Through our dashboard we can educate potential donors on the magnitude of damage that natural disasters can cause. Utilizing interactive visuals provide a digestible way to learn the impact of natural disasters around the world such as the number of deaths, and monetary damage.
EM-DAT – The International Disaster Database Dataset Description: EM-DAT is a catalog of disasters listing detailed information on natural disasters: droughts (famines), earthquakes, epidemics, extreme temperatures, floods, insect infestations, mass movement (dry & wet), storms, volcanos, and wildfires. There is also a data section on technological disasters.
Geographical coverage: Global – country and regional level (primarily cross-country data set, but also contains the name of the sub-national regions affected by disasters) Time span: 1900 onwards. Interactive visuals
- pandas
pip install pandas
- numpy
pip install numpy
- geopandas
pip install geopandas
- geoviews
pip install geoviews
- voila
pip install voila
- holoviews
pip install holoviews
- bokeh
pip install bokeh
- xgboost
pip install xgboost
- Future CPI Prediction Make a CPI prediction of different disasters in the future.
Prediction Obejective: CPI(Community Preparedness Index) Score of how prepared a community is to aid children in a disaster; computed across multiple sectors (hospitals, emergency shelters, child care, etc.).
Dataset: EM-DAT – The International Disaster Database
Model: xgboost
objective = reg:squarederror regression with squared loss
- Earthquake Location Prediction in Next 7 Days
Make a earhtquake location prediction in next 7 days.
Prediction Obejective: Probability of earthquake(magnitude > 2.5) occurrence.
Dataset: USGS Earthquake
Model: xgboost
objective = reg: binary:logistic logistic regression for binary classification, output probability(magnitude > 2.5)
- Future Work
Dataset: Get more data OR use MICE to the missings in EM-DAT dataset(too many missings affect the model performance).
Models: Test with data-insensitive methods (e.g. Bayesian networks) and neural networks.
[Visualization Notebook](https://github.com/srishatagopam/ECE-229-Natural-Disasters-and-Response/blob/main/Group7_Final_notebook.ipynb)Built with Sphinx; preview available here.
Testing done using PyTest, coverage report created using coverage. Preview available here.1.EC2 setup
Create an EC2 instance on Amazon AWS. The "t2 micro" tier is enough for this project and it is free. We add a security group to our instance, then save the "pem" file to the local machine. You can then connect to the ec2 instance and install the above packages through pip or miniconda.
2.Git clone the repository from github to EC2 with:
git clone https://github.com/srishatagopam/ECE-229-Natural-Disasters-and-Response.git
3.Use voila to run the file you want to render as a web page on AWS:
voila Group7_Final_notebook.ipynb
Then you should see a web page on your browser localhost.
4.Connection guide
How to connect to EC2 instance:
How to open jupyter notebook rendered by voila:
Create a private key .pem file, such as the one in our repository
Save this file to a local directory, run the following command:
ssh -i ece229.pem -L 8866:localhost:8866 [email protected] -v -N
Open localhost:8866 on your browser
