Hurricane Path Prediction using a Recurring Neural Network

In this repository you will find the code necessary to train a LSTM-based Neural Network to predict the next position of a hurricane.

The data used was retrieved from the NOAA HURDAT2 database.

As explained on their website it contains the geographic progression data of each hurricane since 1851 (currently am using a version downloaded in 2020, only goes up to 2015)

The hurricanes' coordinates are available for the duration of the hurricane at 4 times in the day : 0000, 0600, 1200, 1800.

I built the model using PyTorch.

It's a very shallow model, as I found out the resultss only worsened when I deepened the LSTM part.

• 1 Input Layer with Batch Normalization

• 1 LSTM Layer with 256 hidden features

• 1 Dense (Fuully Connected Layer)

git clone https://github.com/Zhack47/Hurricane-Path-Prediction.git

cd Hurricane-Path-Prediction/

pip3 install -r requirements.txt

python3 train.py

python3 test.py