CRISP-DM (Cross-Industry Standard Process for Data Mining)
CRISP-DM (Cross-Industry Standard Process for Data Mining) is the most widely used, industry-independent framework for structuring data science, analytics, and machine learning projects. It is cyclical and business-first, ensuring technical work directly aligns with real-world goals.
You can read more about applying the framework through the Data Science PM Guide or explore a breakdown on Towards Data Science.
The 6 Phases of CRISP-DM
• 1. Business Understanding
• Core Focus: Define project objectives, assess the current situation, establish data mining/science goals, and create a project plan.
• Key Question: What real-world business problem are we trying to solve (e.g., reduce churn, boost revenue)?
• 2. Data Understanding
• Core Focus: Collect initial data, describe formats, explore relationships, and verify data quality.
• Key Question: What data do we have, what shape is it in, and are there missing values or errors?
• 3. Data Preparation
• Core Focus: Select, clean, construct, integrate, and format datasets (often the most time-consuming phase).
• Key Question: How do we transform raw data into a clean table ready for modeling?
• 4. Modeling
• Core Focus: Select modeling techniques, design test setups, build algorithms, and assess technical metrics.
• Key Question: Which machine learning or statistical model best fits the patterns in the data?
• 5. Evaluation
• Core Focus: Review the model against business objectives, inspect the process, and decide on next steps.
• Key Question: Does the solution actually solve the original business problem effectively?
• 6. Deployment
• Core Focus: Plan model rollout, set up ongoing monitoring and maintenance, create final reports, and review.
• Key Question: How do we put the model into production and keep track of its performance over time?
Key Characteristics
• Iterative & Cyclical: You rarely move in a strict straight line; teams frequently jump back to earlier phases (such as returning to data preparation or business understanding as new insights emerge).
• Strengths: Universal, non-proprietary, common-sense structure that anchors tech projects in clear business value.
• Weaknesses: Can lean heavy on documentation and originally predated modern cloud-scale big data or agile cross-functional team management (though it can be paired with frameworks like Kanban or Agile).
The evolution of CRISP-DM for Data Science: Methods, processes and frameworks
Publication Date 2024-10-24
1 – The Importance of Domain Knowledge
Data science is often depicted as a field that lies at the intersection of computer science, mathematics/statistics, and domain-specific expertise. Why is domain knowledge important in data science?
No comments:
Post a Comment