1  The CRISP-DM Process

CRISP-DM (Cross-Industry Standard Process for Data Mining) is the most widely used methodology for planning and executing data science and data mining projects. It breaks a project into six phases that form an iterative life cycle — the arrows between phases run in both directions, and the whole loop often repeats as new insight emerges.

1.1 The Six Phases

┌─────────────────────┐      ┌─────────────────────┐
│ 1. Business          │ ───▶ │ 2. Data              │
│    Understanding     │ ◀─── │    Understanding     │
└─────────────────────┘      └──────────┬──────────┘
          ▲                              │
          │                              ▼
          │                   ┌─────────────────────┐
          │                   │ 3. Data Preparation  │
          │                   └──────────┬──────────┘
          │                              │
          │                              ▼
          │                   ┌─────────────────────┐
          │                   │ 4. Modeling          │
          │                   └──────────┬──────────┘
          │                              │
          │                              ▼
┌─────────┴───────────┐      ┌─────────────────────┐
│ 6. Deployment        │ ◀─── │ 5. Evaluation        │
└─────────────────────┘      └─────────────────────┘

1.1.1 1. Business Understanding

Focuses on understanding the project objectives and requirements from a business perspective, then converting this knowledge into a data mining problem definition and a preliminary plan.

• Determine business objectives and success criteria
• Assess the situation (resources, risks, constraints, costs/benefits)
• Define data mining goals
• Produce a project plan

1.1.2 2. Data Understanding

Starts with initial data collection and proceeds to activities that build familiarity with the data, identify quality problems, and form first hypotheses.

• Collect initial data
• Describe data (format, volume, fields)
• Explore data (summary stats, visualizations)
• Verify data quality

1.1.3 3. Data Preparation

Covers all activities to construct the final dataset that will be fed into the modeling tools. Typically the most time-consuming phase.

• Select data (rows and columns)
• Clean data (handle missing values, errors, outliers)
• Construct data (derived attributes, feature engineering)
• Integrate data (merge sources)
• Format data for modeling

1.1.4 4. Modeling

Various modeling techniques are selected and applied, and their parameters are calibrated to optimal values. Often loops back to data preparation.

• Select modeling techniques
• Generate a test design (train/validation/test split)
• Build models
• Assess models (technical performance)

1.1.5 5. Evaluation

Evaluates the model(s) against the business objectives (not just technical metrics) to be certain the model properly achieves the goals before deployment.

• Evaluate results against business success criteria
• Review the process
• Determine next steps (proceed, iterate, or stop)

1.1.6 6. Deployment

Organizes and presents the knowledge so the customer can use it. Ranges from a report to a repeatable, production-grade scoring process.

• Plan deployment
• Plan monitoring and maintenance
• Produce a final report
• Review the project (lessons learned)

1.2 Key Characteristics

• Iterative, not linear — phases feed back into one another; you frequently return to earlier phases as you learn.
• Business-driven — the process begins and ends with business value, not algorithms.
• Tool- and industry-agnostic — applies to any domain and tech stack.
• Documentation-heavy — each phase produces deliverables that inform the next.

1.3 This Repository

This repository is organized as a Quarto book where each CRISP-DM phase is a chapter with its own working folder:

Phase	Folder
1. Business Understanding	01_business_understanding/
2. Data Understanding	02_data_understanding/
3. Data Preparation	03_data_preparation/
4. Modeling	04_modeling/
5. Evaluation	05_evaluation/
6. Deployment	06_deployment/