Data analysis: forecasting, classification and clustering

Содержание и основные понятия

01
Don't confuse the methods.
The projection estimates the future value, the classification selects a pre-known category, the clustering looks for similar groups. All three methods start with quality data and a verifiable target.
Grade 11
PIF · Baseline
Key to programme analysis
02
Route
50.1
Analysis cycle
Question, data, cleaning, model, conclusion.
50.2
Projection
A row, a trend, a horizon, and uncertainty.
50.3
Classification and clusters
Signs, groups and interpretation checks.
50.4
Workshop
Decisions in the data language.
50.5
Outcome
Mini-research and Python route.
03
Limits of methods
Projection
Assessing the future numerical value from past observations, needing time order and verification over a delayed period.
Classification
Selects a famous class by marked examples. A mark-up error teaches the model the wrong rule.
Clusterization
Finds groups without pre-set tags. The result depends on the signs, the scale and the distance measure chosen.
Not an automatic answer
The algorithm only finds the structure in the data submitted; the sample displacement, the target leak and the wrong metric can make a beautiful result useless.