Your own question¶
EPS 88 · PyEarth · final project. Open your own copy on DataHub: click here.
Bring an Earth-science question you really want to answer, find a good dataset for it, and answer it with what you have learned in this class: Python, data science and machine learning. The sections below are an outline, and the ideas under each are only suggestions to help you start. Explore freely: change the sections, try other methods, or go beyond this class if your question needs it. Just say what you did and why. Replace these instructions with your own work.
What you hand in, and when¶
One Gradescope assignment, "Final project: your own question", takes everything.
- Individual work. Start now; Gradescope takes uploads from Wed 25 Nov, and the project is due Wed 16 Dec at 23:59 Pacific time. Mon 7 Dec (RRR week) has no class, but you are welcome to come to the room in class time and work on your notebook.
- Upload this notebook and the data files it reads (each under 50 MB). Upload as often as you like; the last upload is graded.
- Before you upload, run Kernel → Restart Kernel and Run All Cells and check that it finishes without an error.
How it is graded (100 points), by hand:
| Part | Points | What we look for |
|---|---|---|
| The question | 15 | a clear question, why it matters, and what is already known |
| The data | 20 | where it comes from, what it contains, and a first look at it |
| The methods | 30 | several methods from this class, applied correctly, with a reason for each |
| The results | 20 | clear figures and numbers that answer the question |
| The discussion | 10 | how far to trust the answer, what did not work, and what you would do next |
| Runs clean | 5 | the notebook runs from top to bottom without an error |
1. The question¶
What do you want to find out, and why does it matter? What do scientists already know about it? Cite the papers or reports you used.
Ideas: start from something you are curious about, a recent event in the news, or a figure in a paper that you would like to check or extend.
2. The data¶
Where does the data come from, and what does it contain? Load it, look at it, and make a first figure.
Ideas: the data behind a research paper (often in its supplement or on Zenodo or PANGAEA), or a public
archive such as the USGS earthquake catalogue, NOAA climate data or NASA Earthdata. A table you can read
with pd.read_csv is the easiest place to start.
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
3. The methods¶
Apply methods from this class that suit your question, and say why you chose each. Where it helps, compare them with a simple answer that needs no model, and test them on data they did not learn from.
Ideas: fitting a line or a curve, classifying with logistic regression, a forest or an SVM, finding groups with clustering, a neural network, or a bootstrap to see how sure a number is. Use what fits your question; you do not need all of them.
4. The results¶
Show what you found, in figures and numbers. Give each figure labelled axes and a sentence saying what it shows.
Ideas: a map, a time series, a scatter plot of what you predicted against what happened, or a table comparing your methods.
5. Discussion¶
What is your answer to the question? How far should we trust it? What did not work, and what would you try next?