Data Science: Capstone Part 4: Findings + Technical Report

Overview

Data science requires clean data, logical study design, and reproducible results. The best way to do this (and build your portfolio) is to get in the habit of fully documenting your work for your peers and colleagues.

In Part 4 of our Capstone, you’ll assemble a technical notebook that details your model and approach for your peers. It should be written in a straightforward manner, with concisely commented code, documented procedures and reasoning, and logical analysis. Where applicable, include clearly labeled plots, graphs, and other visualizations, explaining any outliers and relationships between features and data.

Start with a brief “executive summary” and then walk us through each portion of your notebokk, step by step. Explain your goals, describe modeling choices, evaluate model performance, and discuss results. Data science reporting is technical, but don’t forget that your approach should tell us a compelling story about your data.

Include any additional code, data, or other materials in appendices, as needed. Above all, your process descriptions should be concise and relevant to your goals.

Goal: Submit a complete notebook of your model for technical stakeholders. This will form the core component of your portfolio.

REQUIRED

1. Begin with an executive summary:
   • What is your goal?
   • What are your metrics?
   • What were your findings?
   • What risks/limitations/assumptions affect these findings?
2. Walk through your model step by step, starting with EDA.
   • What are your variables of interest?
   • What outliers did you remove?
   • What types of data imputation did you perform?
3. Summarize your analysis, including:
   • model selection
   • implementation
   • evaluation
   • prediction/inference
4. Clearly document and label each section
   • Logically organize your information in a persuasive, informative manner.
   • Include notebook headers and subheaders, as well as clearly formatted markdown for all written components.
   • Include graphs/plots/visualizations with clear labels.
   • Comment and explain the purpose of each major section/subsection of your code.
     • Document your code for your future self, as if another person needed to replicate your approach
5. Host your notebook and any other materials in your own public Github Repository.
   • Include a technical appendix, inlcuding links and explanations to any outside libraries or source code used.
   • Host a local copy of your dataset or include a link to a remotely hosted version.

BONUS

1. Describe how this model could be put into production. Consider:
   • How could you continue to validate your model’s performance over time?
   • What steps might you need to take to productionize your model for an enterprise environment?
   • How would you deploy your model publicly? What could you do to setup your model and share it online right now?
2. Create a blog post of at least 1000 words summarizing your approach in a tutorial format and link to it in your notebook.
   • In your tutorial, address a slightly less technical audience; think back to Day 1 of the program - how would you explain and walk through your capstone project to your earlier self?

Deliverable Format & Submission

Format: Jupyter Notebook

Suggested Ways to Get Started

• Use the DSB Data Science Framework to help you organize your information
• For any given step, consider the logic that links it to other steps and clearly desribe each assumption.
• After writing a draft, leave it alone for at least 24 hours. What would you revise, reword, or take out?
• Alternatively, read over a draft with a peer or in a group. What areas do they find confusing or unclear?

Useful Resources

• How to Report Statistics to Technical Audiences
• Data Science Employers Value Research Reports
• What is a good way for a data scientist to construct an online portfolio?

Project Feedback + Evaluation

You will be evaluated on the requirements above using the below rubric.

RUBRIC

| Score | Interpretation | | — | — | | 0 | Project fails to meet the minimum requirements for this item. | | 1 | Project meets the minimum requirements for this item, but falls significantly short of portfolio-ready expectations. | | 2 | Project exceeds the minimum requirements for this item, but falls short of portfolio-ready expectations. | | 3 | Project meets or exceeds portfolio-ready expectations; demonstrates a thorough understanding of every outlined consideration. |

Attached here is a complete rubric for this project.