Final Tutorial

Final Tutorial

In this course you will be responsible for generating a tutorial that will walk users through the entire data science pipeline: data curation, parsing, and management; exploratory data analysis; model building as either hypothesis testing and/or machine learning; and then the curation of a message or messages covering insights learned during the tutorial. Students may choose an application area and dataset(s) that are of interest to them; please feel free to be creative about this!

In the interest of building students’ public portfolios, and in the spirit of “learning by doing”, students will create a self-contained online tutorial to be posted publicly and a 5-8 minute presentation in class. This tutorial will be completed in pairs of students (except for graduate students who will work solo). This assignment will be a publicly-accessible website that provides an end-to-end walk-through of identifying and scraping a specific data source, performing some exploratory analysis, and providing some sort of managerial or operational insight from that data. We will have several milestones associated with the final project including the following. These are each discussed below in more detail.

1. Milestone 1: (20 Points) Identifying a dataset and establishing a GitHub.io Site, Extraction, Transform, and Load (ETL)
2. Milestone 2: (20 Points) Additional Extraction, Transform, and Load (ETL) + Exploratory Data Analysis (EDA)
3. Deliverable 1: (20 Points) In Class Presentation
4. Deliverable 2: (65 Points) Final Website

Important to all coding in this class is becoming a better coder! One handy resource for this is Arie’s Coding Guide. Note that this was written for Intro. to AI CMPS 3140 but many of the same issues apply to this course.

Some examples of high quality notebooks from past Data Science classes:

• Eli and Matt, The Happiness Project
• Addie, Deaths in Louisiana: An Analysis of Available Data
• John (Dung), Airbnb in New York City

All final tutorials from past years can be found on the Past Final Tutorials Page. You should take a look through these, and the topics to get a sense of the scope of these projects.

If you want to see some good examples of Milestone 1, see Example 1 and Example 2. Note that some of the required elements of Milestone 1 have changed so always check the grading rubric to make sure you have all the required elements!

Milestone 1: Groups, Data, Website, and Extraction, Transform, and Load (ETL)

For Milestone 1 you should generate a roughly 1 page writeup (~500 words) listing a partner and one to three datasets that you are considering working with and why. For each dataset you should generate at least one question you hope to answer with that data as well as a small amount of ETL including one interesting stat and one graph. This is just an outline to make sure you are thinking and is not a commitment in any way. This will be published on your GitHub IO page so this also makes sure you’ve figured out how to get it uploaded!

You must also include a short collaboration plan describing how you are working together, what technologies you are using, and when / how often you are meeting to work on this project. Examples include: we setup a private Github repo to coordinate code and we met on Zoom X times…. or even we used LiveShare for CS Code Teletype for Atom or RemoteCollab for Sublime. Failure to turn in a collaboration plan that shows you coordinated will be a loss of professionalism points. The turned in result will need to reflect the understanding of both students

You should load one of these datasets and parse it into shape using the principles of tidy data discussed in class and display the data table in a reasonable format so demonstrating what data you have. This is to show that you have figured out how to get the data into your system and does not need to be a final version, but it should show that you can read in a data source for your project. You should clearly discuss the data and what challenges you had in formatting it.

Possible Sources of Data

This list is just to give you some ideas. Please feel free to scrape websites (legally!) or to find other sources of data that you may find important or interesting.

• Data.Gov: US-centric agriculture, climate, education, energy, finance, health, manufacturing data, and more.
• BigQuery (Google Cloud) public datasets (bikeshare, GitHub, Hacker News, NOAA,…) and many more.
• Kaggle Competition Datasets Billboard Top 100 lyrics, credit card fraud, crime in Chicago, global terrorism, world happiness, etc…
• Amazon AWS Public Data AWS-hosted, various (NASA, a bunch of genome stuff, Google Books n-grams, Multimedia Commons, etc.)
• Data is Plural Lots of interesting and strange datasets in a Google spreadsheet – also see archive here.
• Corgies Datasets Curated set of data from Virginia Tech.
• GitHub Awesome Public Datasets list of datasets that are open and public around the Internet.
• Think Stats Datasets list of interesting datasets from a class very much like this one.
• Data.Nola.Gov list of open datasets from NOLA including lots of financial, voting, etc.
• IPUMS provides census and survey data from around the world integrated across time and space. IPUMS integration and documentation makes it easy to study change, conduct comparative research, merge information across data types, and analyze individuals within family and community contexts.
• Pew Research Center Data Data from the Pew Research Center, mostly public opinion polls of thousands of different things!
• US Census Easy to use data from all the Censuses including many things like names, income, and more! – More specifically https://www.census.gov/programs-surveys/acs
• IRS The IRS publishes all kinds of data including incomes and historical time series.
• The Data Center New Orleans New Orleans specific data of all kinds!
• Five Thirty Eight Data Data from the website FiveThirtyEight – much of this is cleaned and used for their reporting. If you use one of these datasets I expect you to link to an article of theirs using it!
• GapMinder GapMinder curates many public datasets into easy to use formats!
• American National Election Studies Data from US elections, member of congress voting records, and many many more.
• Data USA Another set of curated data, mostly reposts of other things findable from other places but in sometimes easier to use formats.
• CDC Data Download basically everything the CDC publishes.
• US Federal Reserve, FRED Data on everything about the economy that the US Federal Reserve keeps track of.
• US Department of Ag Get data on crop yields, school lunch programs, and many other things from the US department of agriculture.
• Reddit Datasets A subreddit that points to other datasets that may be interesting.
• Police Data Initiative Data
• GitHub of Covid19 Data
• US Cities Survey Data

Github Pages

GitHub provides a service called Pages https://pages.github.com/ that provides website hosting functionality backed by a GitHub-based git repository. We would like you to host your final project on a GitHub Pages project site. To do this, you will need to:

1. Create a GitHub account (or use the one you already have) with your username username. Note that if you want to make it a repository specific webpage instead of your userpage then please go ahead.
2. Create a git repository titled username.github.io; make sure username is the same as whatever you chose for your global GitHub account.
3. Create a project within this repository. This is where you’ll dump your iPython Notebook file and an HTML export of that Notebook file.

These instructions are also given directly on the front page of https://pages.github.com/; following those instructions should be fine!

Note for 2024 CoLab Students: CoLab does not have a “Download as HTML Button” – in order to download your notebook as HTML you need to issue the following command in a cell of your notebook:

%%shell
jupyter nbconvert --to html /PATH/TO/YOUR/NOTEBOOKFILE.ipynb

You will need to replace the path to your notebook file with the path to your actual notebook file. Once you issue that command in CoLab you will see a NOTEBOOKFILE.html file in the same directory that you issue the command in. You will need to download that file, then upload it to the repo that you created on github.io.

In order to make your webpage open up the notebook and make sure everything is as you like it (cells and outputs are showing, no error messages, etc.). Then go to File >> Download As >> and download the notebook as HTML. This will save the HTML files to your disk. You will then need to rename this to index.html and upload it into your github.io page. Once this is done, navigate to username.github.io and make sure the notebook is displayed as you like!

Milestone 1 Submission

You should submit the notebook through Canvas. In the absolute first cell of your notebook you must include your names, project title, and a hyperlink to your webpage at github.io; the webpage must be publicly readable on the internet (i.e, live) and must contain the same work that is in the submitted notebook. That is: the first cell of your notebook must be a markdown cell with a hyperlink to the generated webpage up at yourname.github.io. If this is not correct you will lose points. After this first cell you should continue with the other requirements including a description of your project, links to the data and other relevant resources, a collaboration plan, and the project goals.

If you do this in a group, only one student per group needs to upload the notebook to canvas, but both names should be clearly at the beginning of the notebook.

Milestone 1 Scoring Rubric (20 Points)

If you want to see some good examples of Milestone 1, see Example 1 and Example 2. Note that some of the required elements of Milestone 1 have changed so always check the grading rubric to make sure you have all the required elements!

• (4 Points) Professionalism: You have used both code comments and markdown cells to professionally and clearly document your work including having a clear and clean notebook; linking to resources and documents; and doing so with code that is reasonable and efficient. Your notebook is correct and contains the required links. In addition, you have written code that is interpretable – it contains comments where needed to understand, variable names are reasonable, and code that is reasonable and efficient. You have followed directions to turn in the file, clearly labeling everything. You have cited all sources and how you used them in the written portion of your answers.
• (4 Points) Website: Website is up, link was submitted and is correct. Notebook is professional and clean, the names of the group members, a title for the project, and other good practices as this is publicly posted.
• (4 Points) Project Plan: Project plan is in place, relevant data is identified and links are provided, there are draft questions or hypothesis that the student is going to explore. Plan clearly explains how the data could be used to answer the question and addresses whether or not other data is needed.
• (4 Points) Extraction, Transform, and Load (ETL): At least one data set(s) are loaded correctly using web scraping other techniques. The data is discussed in terms of what it is and how it could be used to answer the question of study. Where the data comes from and how it is collected is clearly documented with links and other relevant details. The data is imported and tidy according to the principles discussed in class. Dtypes are set properly and displayed within the notebook, NaN’s and other techniques are used following best practices discussed in class.
• (4 Points) Exploratory Data Analysis (EDA): You should do some light EDA on your data and show 3 - 5 interesting summary statistics or groups that relate to the questions you are asking of your data along with a brief justification as to why these statistics are interesting and relevant. You should also give at least 1 graphic which shows some interesting property or distribution of your data (skew, histogram, scatter plot) etc., and explain why this is relevant to your question.

Milestone 2: Additional Extraction, Transform, and Load (ETL) + Exploratory Data Analysis (EDA)

Your notebook from Part 1 but expanded. You should remove any data sources or questions that are not being used and you should now have a focused discussion at the top of your notebook on what questions you will attack and a discussion of why the data you are using is correct.

In addition you should also have at least 5 graphs and summary statistics/tables including cross-tabulations, groups, measures of central tendency, and/or any of the other EDA techniques discussed showing that you have started to investigate your data and have an understanding of what is in your dataset. These graphs should be documented much as we did in Project 1 where you describe what the graph is showing and why. Use of appropriate Markdown cells is essential.

At the end of your notebook you should outline at least two ideas of what model you will be building to in the rest of your notebook. Example ideas include “We plan to use the time and date variable to predict parking violations in a zip code” or “We plan to test if there is a statistically significant positive correlation between income and happiness.” The graphs and summary statistics should inform what questions you have decided to ask here, if they don’t, compute more summary statistics!

Milestone 2 Submission

You should submit the notebook through Canvas. The first cell of your notebook in the markdown must be a hyperlink in your notebook to the webpage at github.io, and the webpage must be publicly readable on the internet (i.e, live). That is: the first cell of your notebook must be a markdown cell with a hyperlink to the generated webpage up at yourname.github.io. If this is not correct you will lose points.

If you do this in a group, only one student per group needs to upload the notebook to canvas, but both names should be clearly at the beginning of the notebook.

Note: If the dataset you are using is not available from a website then it must also be uploaded to your GitHub site. If I cannot download the data to verify the code you have submitted you will lose points.

Milestone 2 Scoring Rubric (20 Points)

• (5 Points) Professionalism: You have used both code comments and markdown cells to professionally and clearly document your work including having a clear and clean notebook; linking to resources and documents; and doing so with code that is reasonable and efficient. Your notebook is correct and contains the required links. In addition, you have written code that is interpretable – it contains comments where needed to understand, variable names are reasonable, and code that is reasonable and efficient. You have followed directions to turn in the file, clearly labeling everything. You have cited all sources and how you used them in the written portion of your answers.
• (5 Points) Extraction, Transform, and Load (ETL): You are now using more than one dataset and they are both loaded correctly using web scraping other techniques. The data is discussed in terms of what it is and how it could be used to answer the question of study. Where the data comes from and how it is collected is clearly documented with links and other relevant details. The data is imported and tidy according to the principles discussed in class. Dtypes are set properly and displayed within the notebook, NaN’s and other techniques are used following best practices discussed in class.
• (5 Points) Exploratory Data Analysis (EDA): At least 5 graphs and/or summary statistics/tables including cross-tabulations, groups, measures of central tendency, and/or any of the other EDA techniques discussed showing that you have started to investigate your data and have an understanding of what is in your dataset. Each of these pieces of analysis should have a corresponding markdown cell that clearly explains what is being shown and why.
• (5 Points) Model Questions: You have identified at least two plausible models that you could build using your data and the graphs in your EDA section support continuing the investigation. You have clearly identified independent and dependent variables and appropriate test/prediction methodology for each question.

Deliverable 1: Final Presentation

You (and your partner) will give a 4-5 minute presentation in class to your peers. In this presentation you are role playing a bit – your job is to convince the audience that the problem you investigated was interesting and meaningful. An ideal talk will cover every step of the Data Lifecycle that we have discussed in class: (1) an introduction of the problem and what question you want to solve, (2) where you got your data as well as briefly how you loaded and preprocessed your data, (3) some exploratory analysis to show important features, (4) a hypothesis about the data and supporting evidence, (5) an important conclusion to draw from your data. Your presentation should only consist of 5-7 slides – one for each step!!

Some resources for giving a good talk:

1. Simon Peyton Jones, How To Give a Great Research Talk
2. Matt Might, 10 Tips for Academic Talks

Here are some past short decks that are very good:

1. Sam Bruchhaus Saints Third Down Analysis
2. Nick Ouellet and Garrett Gilliom Education in LA and NJ
3. Ila Keshishian and Bennett Kahn NOLA Court Docket and Police Analysis
4. Zachary Airington It’s all Fake News: Big 5 Personality Types and Conspiracy

Deliverable 1 Scoring Rubric (20 Points)

• (4 Points) Professionalism: Are the slides well constructed? Are the graphs labeled and clear? Did you speak clearly and have a plan for the discussion?
• (4 Points) Organization: Was the talk clearly organized? Did you explain where your data came from, why the question was important, and what the outcomes are? You should address each stage of the DataScience LifeCycle in this talk.
• (4 Points) Data and Motivation: Does the talk make the reader believe the topic is important (a) in general and (b) with respect to data science? Does the data you selected allow you to answer the question you set out for? Is there a clear and explainable question being answered by this data?
• (4 Points) Model and Science: Did you clearly explain why you performed the analyses you chose? Did you adequately defend any assumptions you made for your model? Does your analysis support your conclusions?
• (4 Points) Peer Evaluation: Were you present in class both days during presentations, did you provide feedback to the other students on their projects and ask questions on the message board. Were you a good peer?

Schedule

Note: You must upload your slides through Canvas at least 2 hours before class time on the day of your presentation. Failure to do this will result in a minimum of all points lost for Professionalism. You will be required to be in person to present unless you have made other arrangements with me. I will compile all slides into one deck for the individual presentation days.

Deliverable 1 Submission

You should submit your slides (Powerpoint or PDF only) through Canvas at least 2 hours before class time on the day of your presentation. Both students in groups should submit the same set of slides. If you do not sign in for class these days and engage with the discussion you will not receive points for the Peer Evaluation section.

If you do this in a group, only one student per group needs to upload the slides to canvas, but both names should be clearly at the beginning of the notebook.

Deliverable 2: Final Tutorial

In general, the tutorial should contain at least 1500 words of prose and 150 lines of (non-padded, legitimate) Python code, along with appropriate documentation, visualization, and links to any external information that might help the reader. The big difference between Milestone 2 and the final is the inclusion of your model and testing – your final notebook should clearly identify the model and then use that model to support conclusions and inform the construction of the rest of the notebook.

You should make sure the top of your notebook is well organized and provides a clear motivation for the questions. You may need to rewrite significant parts of what you have done to accomplish this but you should consider the final notebook to be a portfolio piece! Remove things that are no longer true or necessary!

Final Notebook Scoring Rubric (65 Points)

• (5 Points) Motivation: Does the tutorial make the reader believe the topic is important (a) in general and (b) with respect to data science? Is there a clear, coherent argument for this at the beginning of the notebook? Is the model question well motivated.
• (5 Points) Understanding: After reading the tutorial, does the reader understand the topic?
• (5 Points) Subjective Evaluation: If somebody linked to this tutorial from Hacker News, Reddit, Towards Data Science, or some other general interest website, would people actually read the whole thing?
• (5 Points) Further Resources: Does the tutorial “call out” and specifically provide links to other resources that would help the reader understand basic concepts, deep dive, related work, etc?
• (5 Points) Prose/Organization/Professionalism: Does the prose in the Markdown portion of the .ipynb add to the reader’s understanding of the tutorial? Are the graphs labeled and clear? Is the writing clear and explained well. Was the notebook clearly organized? Did you explain where your data came from, why the question was important, and what the outcomes are? There should be a section heading for each step of the Data Science LifeCycle.
• (10 Points) Data: Is the data selected appropriate? is it loaded in a coherent way? clean and Tidy? Dtypes set? are there issues in the data and if so are they explained? is enough data present and if not, why not? Is the data manipulated properly and coherently?
• (10 Points) Code: Does the code help solidify understanding, is it well documented, and does it include helpful examples? Is the code correct? Did you show that the results of intermediate steps are correct in the tutorial? Is there enough code to be complex enough to require a tutorial?
• (10 Points) Exploration and Analysis: Is the data explored and explained in a coherent way, are there graphs, tables, summary statistics, and other methods of analysis present that let the user understand the data in a coherent and clear way? Are these graphs well labeled, clear, and show trends/information of interest? Note that just throwing 20 graphs in is not interesting, if a graph or piece of analysis is not answering a question, it should be removed.
• (10 Points) Science/Model: Did you clearly explain why you built the model you chose? Did you adequately defend any assumptions you made? Does your analysis support your conclusions? Is the model correct and follow proper data science techniques?

Final Notebook Submission

You should submit the notebook through Canvas. The first cell of your notebook in the markdown must be a clickable link to the webpage and the webpage must be live. That is: the first cell of your notebook must be a markdown cell with a hyperlink to the generated webpage up at yourname.github.io. If this is not correct you will lose points.

If you do this in a group, only one student per group needs to upload the notebook to canvas, but both names should be clearly at the beginning of the notebook.

Note: If the dataset you are using is not available from a website then it must also be uploaded to your GitHub site. If I cannot download the data to verify the code you have submitted you will lose points.