20. • Define business
objectives
• Translate business
objectives into ML
objectives
• Collect and verify
data
• Assess the project
feasibility
• Create POC
CRISP-ML(Q) Phase
•
• Business & Data
• Feature Selection
• Data Selection
• Class Balancing
• Cleaning Data
• Feature Engineering
• Data Augmentation
• Data Standardization
•
• Data Engineering
• Define quality
measure of the
model
• ML Algorithm
selection
• Domain Knowledge
• Model Training
• Transfer Learning
• Model Compression
• Ensemble Learning
• Model
Documentation
•
• ML Model Engineering
• Validate model’s
performance
• Determine
robustness
• Increase model’s
explain ability
• Model deployment
• Model evaluation
documentation
•
• ML Model Evaluation
Agile does not make sense for data science in academic settings.
Agile does not make sense for data science in academic settings.
The Agile Data Science process will take place at this level of the data value pyramid. Therefore, the other foundational items must be set before implementing.