Global temperature prediction with backpropagation algorithm

Inspiration

World climate/temperature prediction suffers several pains. The past data is limited, influencial human events are unpredictable (ed. industrial revolution), and landscape can have huge impact on the model.etc. We can take care of those factors by very careful observation at different locations and make different prediction models which often produce different result ---- of which only one can be correct. Therefore, we believe an AI model will be better for making the prediction.

An accurate climate/temperature prediction can be helpful to agriculture, ecology studies, and many other aspects.

What it does

We utilized a feedforward neural network and created an AI trained with backpropagation algorithm. The input of this neural network is CO2-equivalent emission, population, and (difference in ) world temperature anomaly. The output is annual world temperature change. We also created some classical world temperature anomaly prediction models, and we compared the validity/accuracy between two models (the AI and non-AI). Besides, we created a personal CO2E emission estimator. With this module, you can directly feel your contribution to climate change.

Challenges I ran into

We had issue with selecting the input/output of the neural network since there were no "real value" for predictions. We eventually decided to utilize less data for traning set, and use known data as our target.

Lack of training set.

Accomplishments that I'm proud of

We made the algorithm work! According to our AI model, temperaure will decrease next year. Thanks to human.

What I learned

How backpropagation algorithm works. The threats our world is facing. How intense Hackathons are.

What's next for Global temperature prediction with backpropagation algorithm

A more accurate model can be trained with more data. We should specify the constraints for the model. For example, landscape, location, human activity index. etc. There's a huge paper by NOAA about how to make climate predictions, please read if you are interested.