Data Science

Context: FIT1043_MOC · extracting knowledge/value from data across its lifecycle · realised through the standard value chain · sits beside Machine Learning and related fields

Quick Revision

  • 🎯 Objective: define data science as the extraction of knowledge/value from data ➔ the broad, lifecycle-wide view, not just “ML on big data”.
  • 📦 Core Components: hacking skills ∩ maths & stats ∩ substantive expertise ➔ the Conway Venn intersection.
  • ⚡ Key Constraint: the danger zone — skills + domain but no maths/stats ➔ plausible-looking analysis with no rigour.

📝 Core

1. Defining It (the spectrum)

  • Circular (avoid) ➔ “what a data scientist does” ➔ practically useless; never answer this way.
  • Better ➔ “the technology of handling and extracting value from data”.
  • Narrow ➔ “machine learning on big data” ➔ useful but too limited.
  • Broad (unit view) ➔ inter-disciplinary use of scientific methods/algorithms to extract knowledge from structured and unstructured data across the whole lifecycle.
  • Hal Varian ➔ take data and understand → process → extract value → visualise → communicate it — a hugely important skill.

2. Drew Conway’s Venn Diagram

  • Three circleshacking skills, maths & statistics, substantive (domain) expertise.
  • Centre ➔ all three overlap = data science.
  • Applied ➔ Microsoft traffic forecasting, iOS predictive text, Google Translate, Amazon recommender, health/saturated-fat studies.

⚖️ Core Decision Matrix

OverlapRegionNote
hacking ∩ maths/statsMachine Learningmany CS grads start here (ML engineer)
maths/stats ∩ domain expertiseTraditional researchdeep domain, little technology
hacking ∩ domain expertiseDanger zoneanalysis looks valid but lacks rigour
all threeData sciencethe target combination

When It Flips: the danger zone is dangerous precisely because it omits maths/stats — results appear legitimate with no understanding of how they were produced; data science needs all three skill sets together.

🧠 Active Recall