These are other ways you can earn innovative besides adding onto previous assigments.
Tutorial¶
Write a notebook that explains a concept related to a skill with examples in a real dataset and with visuals or a toy dataset (minimal number of columns rows)
Cheatsheet¶
Make a detailed reference with code outputs on a topic or a few topics.
Blog post¶
Write a blog post styled Notebook that compares or analyzes something, for example:
how do different ways of loading data compare
describe best practices you’ve learned and show why they’re good with examples
summarize what you learned from a guest speaker
Practice Problems and Solutions¶
Write practice problems that build off of the lecture notes. Include solutions and descriptions for each. These can be open ended or multiple choice questions with plausible distractors. A plausible distractor is an incorrect answer that represents a way that you think someone could misunderstand.
For example if the question is 37 + 15 = ?, MCQ with plausible distractors might be:
52 (correct)
412 (didn’t carry the one, correctly: 7+5 = 12, 3+1 = 4)
42 (dropped the one 7+5 = 12, ones place is 2, 3+1 = 4)
43 (carried one into wrong column, 7 + 5 = 12, 1+2 = 3, 3+1 = 3)
Long single analysis¶
Collect data from multiple sources, prepare each for analysis, and merge them together then do some exploratory data analysis. Describe each step, interpret all outputs, and put the analysis in context of the Data Science Process.
This would be one long notebook that covers many skills at once.
Create datasets that fail¶
Create datasets that violate assumptions of a model we have learned. The sklearn data generators are a good place to start.
Build a data set for Prediction¶
Build a dataset that works for prediction (classification, regression, or clustering) from other sources.
Organize your knowledge¶
Develop some sort of visual aid that demonstrates how you understand some aspect(s) of data science working. Think of this as something that future students could use to help them learning, so assume prior knowledge topics covered earlier than the one you are demonstrating.
This could be a concept map, a table that shows how you’ve traced how something works or any other sort of conceptual tol that helps convey your understanding.
Try alternative libraries/ tools¶
One option for workflow level 3 is to use other data science skills and reflect on how what we have learned so far helped you learn a new set of tool as an alternative way to do things.
Try feature engineering or representation learning¶
Try different transformations and see how they impact how well a model performs.
This could be using sklearn.feature_extraction tools or trying different
types of neural network layers at the beginning.