# AI for Understanding Earthquakes — Data

来源：https://www.kaggle.com/competitions/earthquake-prediction/data

本地 Meta Kaggle 快照；不是本次实时页面。

# Dataset Description

The data for this competition was recorded continuously and then split sequentially as ``train`` and ``test`` data. It contains all the information necessary to complete both parts of the challenge.

## **File Descriptions**
**train.csv:** Data you can use to train your models. Includes features and label data.
**test.csv:** Acoustic emission measurements you should use to generate your predictions.
**sample_submission_long.csv:** Example of a properly formatted submission file.
## **Data Fields**
**time_stamp:** time at which acoustic emission, time-to-failure, and end-point displacement measurements were taken
**ae_1 through ae_6000:** acoustic emission signal
**ttf:** time-to-failure reading
**EP_disp:** end-point displacement measurement
**EP_disp_FUT_1:** end-point displacement measurement for the subsequent timestep
**EP_disp_FUT_2:** end-point displacement measurement two timesteps in the future
**EP_disp_FUT_3:** end-point displacement measurement three timesteps in the future

In ``train.csv``, the first column is the ``time_stamp`` at which the measurements in the corresponding row were taken. The next 6000 columns consist of acoustic emission readings. The goal is to predict the final five columns ``ttf``, ``EP_disp``, ``EP_disp_FUT_1``, ``EP_disp_FUT_2``, and ``EP_disp_FUT_3``.

In ``test.csv``, we just provide timestamps and the acoustic emission readings. From this data, your models should predict the ttf as well as displacement values for the current timestep and the next three future timesteps.

Data is collected in Chris Marone's rock physics lab at Penn State (https://personal.ems.psu.edu/~cjm38/).