This is one of the two main lessons of the module. We aim to go through all the stages of a typical data science project, emphasising the scoping/ideation/translation parts at the beginning. We want to give multiple examples of real projects to explain hurdles that are typical and expected and how to best manage them. We also want to show the iterative nature of scoping and of the whole life cycle. Some of the material will overlap with Turing Commons lifecycle material.
Translating a research question into a data science task.
Ambiguity and complexity in scoping with some real case studies.
MVPs
Collaborating with clients, adoption
Getting and wrangling data
Feature engineering, selection
Model training and evaluation
Production
Monitoring performance and updating
Handover
Some examples we can use
Examples of real projects where the scope was not clear and needed clarification and how it was achieved, e.g. some examples from the UA challenge area (maybe without explicitly mentioning them). The role of a DS in this to translate, make the possibilities visible to non-technical collaborators.
Examples where adoption can be an issue due to norms/culture/skepticism
Production issues in existing projects
Models that deteriorated over time (I think there was a project with the topic of monitoring performance when data shift/change over time).
Module 1.2: Project life cycle
Description
This is one of the two main lessons of the module. We aim to go through all the stages of a typical data science project, emphasising the scoping/ideation/translation parts at the beginning. We want to give multiple examples of real projects to explain hurdles that are typical and expected and how to best manage them. We also want to show the iterative nature of scoping and of the whole life cycle. Some of the material will overlap with Turing Commons lifecycle material.
Outline of this section
Some examples we can use
Duration
60 minutes
Estimate time for developing/writing the section
20 hours