Project Zeus| Automated Detection of Hazards
Awards & Nominations
Project Zeus has received the following awards and nominations. Way to go!
Automated Detection of Hazards
Hurricane Predictor Web App
Summary
*UPDATE* Since the honorable judges had asked us to have a web app as well our team has designed a simple and seamless front end web app for our judges :).APP LINK POSTED BELOWThis web app shows the data used and analysis done to determine the target for our model. The model can predict the intensity of the hurricane with high accuracy. The team has spent a lot of time creating wonderful data visualizations to pick the target and training variables.Please click the 'Let us analyse the information' checkbox to know more.
How We Addressed This Challenge
We have developed a web app that runs a machine learning model for predicting hurricanes.
Due to the fluctuating climate change recently, we believe it is important to have quick and simple algorithms that can tell the intensity of an approaching natural disaster such as a hurricane before it even occurs. This can save a lot of lives and a lot of expensive national resources.
Our model does exactly that. This is not a vision or a fantasy project. We have built a live model that can even be tested by the user on the web app.
(app link: https://project-zeus-space-apps.herokuapp.com/)
The web app has an About box which when selected shows important information about the app.
following that we have the checkbox to give a summary of the data that we analysed. This gives additional information and a very useful insight to the user of the webapp.
Finally we have the section that runs the model and shows the accuracy with which our model is able to predict. After this we have a 'Try it Yourself' button which when clicked will generate a random value for the parameters in the dataset, show the user the expected output as well as what the model predicted.
The model would be able to predict the right answer almost all the time. This shows its reliability in the commercial world. The web app is the front-end of a very advanced and dynamic machine learned model that our team has made from scratch. It uses DecisionTreeRegressor to provide the most accurate results.
Through our web app, we hope to achieve a greater accuracy in the world of machine learning as this can truly work for the greater good of mankind
How We Developed This Project
Natural disasters occurances are becoming greater and greater. In 2020 we saw the 2020 Atlantic hurricane season, an ongoing tropical cyclone season which has featured tropical cyclone formation at an unprecedented rate. This has caused billions of dollars in damage, especially in North American coastal cities. Hence we felt the strong need to create a better detection model to prepare authorities and weather data scientists on possible storms.
We first went through the provided sources by NASA SpaceApps and found the NOAA HURDAT Dataset to be suitable for our project. After obtaining the data we plotted correlation matrices and saw the importance each feature. We had a very long discussion related to how the data correlates to storm formation. After getting some insights we decided the categorical feature to use as our label. Then we proceeded with the code of our algorithm
We used data in the form of .csv files for the project. We used the pandas library in python to load the dataframe. We used the Seaborn and Matplotlib libraries to visualize our data and represent our model. We trained our model using a DecisionTreeRegressor from SciKit learn. After creating the model we created an interactive web app using streamlit library and hosted using Heroku.
We had difficulty as the data had a lot of missing values. We were able to clean the data to a high level before training our model. In addition choosing an algorithm to train our model was a tough choice so we tried various different algorithms until we found the one best suited for it. Finally creating a front end from scratch was a challenge but with inputs from the judges we were able to create a =n interactive effective web application
How We Used Space Agency Data in This Project
We collected all the information we needed from the resources section of the challenge or from what the challenge moderators provided us in the spaceapps chat.
We used the Atlantic dataset (known as Atlantic HURDAT2) from National Oceanic and Atmospheric Administration (NOAAA) for our machine learning models. The data has a comma-delimited, text format with six-hourly information on the location, maximum winds, central pressure, and (beginning in 2004) size of all known tropical cyclones and subtropical cyclones. The data wasn’t maintained very well before 2004 because it didn’t have the wind speed and pressure data.
We had to run cleaning algorithms on the data after 2004 like converting the latitude, longitudes and dates to correct usable format. We ran a co-relation matrix algorithm on the data and we found that the wind speeds and pressure were very closely related. Also this data was very well labelled with status of storms that happened dividing it between difference categories of tropical, subtropical and extra-tropical storms that helped us to train our regression model to output the predicted label
Project Demo
APP LINK: https://project-zeus-space-apps.herokuapp.com/
PPT EXPLAINING WORKING: https://drive.google.com/file/d/16iJPhO-2vt_iuH-VxEKCydGUAzJ9wxMj/view?usp=sharing
Project Code
Data & Resources
source: https://www.nhc.noaa.gov/data/
Atlantic Data: https://www.nhc.noaa.gov/data/hurdat/hurdat2-1851-2019-052520.txt
Pacific Data: https://www.nhc.noaa.gov/data/hurdat/hurdat2-nepac-1949-2019-042320.txt
The Atlantic dataset (known as Atlantic HURDAT2) has a comma-delimited, text format with six-hourly information on the location, maximum winds, central pressure, and (beginning in 2004) size of all known tropical cyclones and subtropical cyclones.
The Pacific dataset was provided on 23 April 2020 to include the best track for 2018's Walaka in the Central Pacific. The best track provided in this database for the Erick (EP062019) in the Central Pacific Hurricane Center's area of responsibility (between 140 and 180W longitude) is an operational estimate and has not yet been post-storm analyzed. The final best track for this system will be updated when it becomes available.
The status of system labels we are looking at are:
HU – Tropical cyclone of hurricane intensity (> 64 knots)
TS – Tropical cyclone of tropical storm intensity (34-63 knots)
TD – Tropical cyclone of tropical depression intensity (< 34 knots)
EX – Extratropical cyclone (of any intensity)