C01 - Surf Forecast

• 

  train and validation accuracy

Inspiration

Inspiration of our 3 team members:

1. direct connection to weather (also at the sea) on a daily basis
2. Interest in surfing
3. Working at SAR as a sea rescuer
4. Studying ocean science

What it does

Our Neural Network decides based on wind and temperature data from one day, if it is a good day for surfing or not

How I built it

We build a small feed forward network (FFN) to fit our data. We used github, google colab, keras, tensorflow, numpy, csv, relu and sigmoid as activation functions.

Challenges I ran into

Our team started to learn ML / DL some months ago and is building up experiences of data science methods and coding in new environment with python and the necessary libraries. The data preparation cost us most of the time: processing the raw csv files, work around missing days and different amount of data on different days and in the end feed our neural network with the correct data was a very tricky challenge for us.

Accomplishments that I'm proud of

We build a network that straight away delivered decent results and it was great to became a successful team in such a short time of 2 days.

What I learned

We learned a lot about data preparation / preprocessing, applied machine learning with tensor flow and python, usefull libraries. We discussed about different types of NN architectures, activation functions, online Hackathons.

What's next for C01 - Surf Forecast

Tuning the networks parameters and try different architectures like RNN (Recurrent Neural Network with LSTM), try other spots (we focused on Kiel Leuchtturm weather data), create forecasts for different watersport activities like surfing, windsurfing, kite surfing , sailing, SUP.

And grazie mille to Luca!!! He is doing our Deeplearnig Course and while working on his own project, he helped us out with the data processing.

Built With

• keras
• matplotlib
• numpy
• python
• tensorflow