High-Level Project Summary
The project goal is to utilize a pre-existing library called prophet to do a forecasting time series from the wind ion parameter data. This method helps to make pre-determined decisions that can save the earth from possible blackouts in the far and near future. This idea may not be the most suitable one to predict future disasters yet it is the best we have to estimate the danger of such events.
Link to Final Project
Link to Project "Demo"
Detailed Project Description
The challenge's main problem and goal is to develop a machine learning algorithm to correctly track changes in the peak solar wind speed and provide an early warning of the next potential Carrington -like event. Data on wind and temperature and density and wind speed is available and archived at NASA’s online database and can be accessed directly from the database’s website. The solution we proposed is to first do Exploratory data analysis (EDA) on the dataset of wind and DSCOVR we retrieved from the NASA Website, where we graph the data and find periods containing clear qualitative differences between the two data sets, and then we remove them,
Space Agency Data
We used the wind ion parameters provided to us by NASA. We converted the data type from cdf to csv so that we can read the data and analyze it. We then printed the data and chose the required parameters for the testing which we noticed by comparing the acronyms of the ions with the data. In the next steps, we fetched the data from a designated dictionary for every ion.
Hackathon Journey
The experience we had in this hackathon was truly incredible and we are extremely glad that we were able to take part in such a competition. This experience gave us a sense of purpose when we realized the effect of the models in minimizing the risk of such global disasters. To illustrate, these models can save the world $2 trillion which is even worse than the global financial crisis. Our allure to both physics and machine learning inspired us to choose this challenge in particular. This challenge was able to feed our insatiable curiosity about these fields in particular.
References
1- Larsen, E. (2022). MACHINE LEARNING MODEL SURVEY WITH THE DATASET FOR SOLAR FLARE PREDICTION. Peopletec. Retrieved from https://arxiv.org/ftp/arxiv/papers/2110/2110.07658.pdf
2- Smart, D., Shea, M., & McCracken, K. (2006). The Carrington event: Possible solar proton intensity–time profile. Advances In Space Research, 38(2), 215-225. doi: 10.1016/j.asr.2005.04.116
3-Taylor, S., & Letham, B. (2017). Forecasting at scale. doi: 10.7287/peerj.preprints.3190v2
4- Wind ion parameters from NASA