This is my implementation of the LANL Earthquake-Prediction challenge The goal of this project is to predict upcoming earthquakes based on a stream of seismic activity. The training data consists of a ~10GB large csv file which contains a seismic_activity and a time_to_failure column, which represents how many seconds are left before the next earthquake:
|seismic_activity (V)||time_to_failure (s)|
The goal of the challenge is to find the "best" correlation between a range of seismic activities and the time remaining before the next earthquake, so that we can predict the time_to_failure column given only the seismic_activity column.
In the training data, we can observe an earthquake every time the time_to_failure values jumps from zero up to a higher value, giving us 16 earthquakes in total:
To download the training/test data, please check out the link mentioned above.
Read more about this challenge on my blog.