Data analysis cycle

Содержание и основные понятия

01
Six steps
1. Question
"Is the average time of journey changing by season?" is measurable.
2. Data.
Determine the source, period, units and permissible values.
3. Cleaning.
Lock down the pass and anomaly processing.
4. Method.
Choose a summary, a model, a classification or clusters.
5. Verification.
Compare the result with known data or a delayed part of the set.
6. Conclusion
Show me the limit and don't turn the connection into a proven cause.
02
Ethics
Risk
Don't use sensitive personal data to evaluate a person or make a decision without a person.
03
Analysis of deviations
Deviation
The difference is between the value observed and the value expected. Compare it with the units, the usual dispersion and context, and not declare any difference as a mistake.
Step
Question
Baseline
Plan, medium, median, past or forecast?
Scale
Absolute deviation or percentage relative to base?
Reason
Error input, seasonality, process change, or rare event?
Decision
Correct the data, see further, or change the process?