Chapter 1 — What Can We Do With Data?¶
1. What is Data Analytics?¶
- Core idea: data → information → knowledge
- Analytics = the science of analyzing raw data to extract useful, valid, human-understandable patterns
2. What Is Data?¶
| Term | Definition |
|---|---|
| Instance (object/row) | One example of the concept being studied |
| Attribute (feature/column) | A characteristic of an instance |
| Attribute value | The number or symbol assigned to an attribute for a specific object |
A measurement scale is the rule (function) that maps an attribute to a value.
3. Four Types of Attributes (Stevens' Taxonomy)¶
| Type | Properties | Operations | Examples |
|---|---|---|---|
| Nominal | Distinctness only (=, ≠) |
Mode, entropy, χ² | Eye color, zip code, ID |
| Ordinal | Distinctness + Order (<, >) |
Median, percentiles, rank corr. | Grades, {good/better/best}, mineral hardness |
| Interval | + Meaningful differences (+, −) |
Mean, std dev, Pearson's r | Temperature (°C/°F), calendar dates |
| Ratio | + Meaningful ratios (×, ÷), true zero |
Geometric/harmonic mean, % variation | Temperature (K), age, length, counts |
Categorical vs. Numerical¶
Discrete vs. Continuous¶
- Discrete — finite or countably infinite values (integer variables, binary is a special case)
→ Can be Nominal, Ordinal, Interval, or Ratio - Continuous — real-valued (floating-point); e.g., temperature, height, weight
→ Interval or Ratio only
4. Data Structures¶
| Structure | Core idea | Example |
|---|---|---|
| Record | Fixed attribute set per record; default tabular form | spreadsheet rows |
| Data Matrix | Record data where all attributes are numeric → m × n matrix, each row a point in n-D space | gene expression table |
| Document | Each doc → term-frequency vector (attributes = unique terms, values = counts) | bag-of-words corpus |
| Transaction | Each record = variable-length set of items; no fixed schema | shopping basket |
| Graph | Objects = nodes, relationships = edges | social network, molecule |
| Ordered | Sequential (item sets over time), Genomic (DNA over {A,T,C,G}), Spatio-Temporal (location + time) | clickstream, weather map |
5. Taxonomy of Analytics¶
| Type | Question Answered | Output |
|---|---|---|
| Descriptive | "What happened?" | Summaries, patterns (applied directly to data) |
| Diagnostic | "Why did it happen?" | Root cause analysis |
| Predictive | "What will happen?" | Model — a generalization used for future predictions |
| Prescriptive | "What should we do?" | Recommended actions/strategies |
A model in predictive analytics is a generalization learned from data that can generate predictions on new, unseen instances.
6. Data Analytics Methodologies¶
CRISP-DM (Cross-Industry Standard Process for Data Mining) — 6 phases¶
Business Understanding → Data Understanding → Data Preparation
↑ ↓
Deployment ← Evaluation ← Modeling
Each phase has sub-tasks (e.g., Data Preparation includes: Select, Clean, Construct, Integrate, Format data).
KDD Process (Knowledge Discovery in Databases) — 9 steps¶
- Learn application domain
- Create target dataset
- Clean & pre-process (missing values, noise, format)
- Reduce & project (feature selection / engineering)
- Choose mining function (clustering? classification? regression?)
- Choose algorithm
- Apply data mining
- Interpret results → loop back if needed
- Deploy discovered knowledge
Mining function → analytics branch: - Summarization & Clustering → Descriptive - Classification & Regression → Predictive
7. Quick-Reference: Key Distinctions¶
| Concept | Distinction |
|---|---|
| Tabular vs. Relational data | Relational uses foreign keys to express relationships between records across multiple tables |
| Attribute vs. Attribute Value | Same attribute can have different scales (feet vs. meters); same value set can represent different attributes |
| Algorithm vs. Method | Method = what to achieve; Algorithm = how, step-by-step, implementable in code |
Apply the Chapter¶
Continue with the matching Chapter 01 exercises, take the Chapter 01 quiz, or build one of these focused projects:
- Attribute Auditor — infer each column's likely measurement scale, flag discrete vs. continuous values, and report missingness.
- Dataset Type Explorer — classify a dataset as record, matrix, transaction, graph, or ordered data.
- Attribute Scale Exercises — solve the measurement-scale examples in both Python and R.