Data Science Process (Standard Value Chain)

Context: FIT1043_MOC · how a Data Science project actually runs · the unit’s model of the lifecycle · many tasks, not all labelled “data science”

Quick Revision

  • 🎯 Objective: run a project as an ordered pipeline ➔ Collection → Engineering → Governance → Wrangling → Analysis → Visualisation → Operationalize.
  • 📦 Core Components: Wrangling ➔ clean/pre-process | Analysis ➔ learning/discovery | Operationalize ➔ turn results into value.
  • ⚡ Key Constraint: a data scientist is familiar with most stages but need not be expert in all; some stages come from other fields (engineering, business).

📝 Core

1. The Standard Value Chain (unit model)

#StageWhat it covers
1Collectiongetting the data
2Engineeringstorage + computation across the full lifecycle
3Governanceoverall management of data across the lifecycle
4Wranglingdata pre-processing and cleaning
5Analysisdiscovery — learning, statistics, visualisation
6Visualisationarguing the results are significant and useful
7Operationalizeputting results to work to gain value

2. Illustrating the Process (10 tasks)

  • Front end ➔ pitching ideas → collecting data → integration (many sources) → interpretation (a schema).
  • Back end ➔ governance (caring for data + subjects; standards/formats) → engineering (back-end works).
  • Core loop ➔ wrangling (inspect/clean) → modelling (conceptual/mathematical/functional; ML/stats) → visualisation → operationalize.
  • Roles ➔ different data scientists carry different mixes of skills and domain knowledge (see Data Science vs Related Disciplines).

⚠️ Common Mistakes

  • 💡 Wrangling ≠ Analysis ➔ wrangling is cleaning/pre-processing to make data usable; analysis is the discovery/modelling that extracts insight — do not conflate them.
  • 💡 Not every stage is “data science” ➔ collection/engineering/governance draw on computer engineering, business, and data management; a data scientist need not master all.

🧠 Active Recall