Ensuring the Robustness and Reliability of Data-Driven Knowledge Discovery Models in Production and Manufacturing.

Area of Science:

• Data Science
• Machine Learning
• Knowledge Discovery

Background:

• The Cross-Industry Standard Process for Data Mining (CRISP-DM) is a key framework for data analytics and machine learning in production and manufacturing.
• Practical implementation of CRISP-DM faces challenges with data and model development, requiring industry-specific solutions.

Purpose of the Study:

• To review the CRISP-DM framework and its extensions.
• To introduce a novel Generalized Cross-Industry Standard Process for Data Science (GCRISP-DS) framework.
• To enhance the flexibility and robustness of data science processes.

Main Methods:

• Detailed review of the CRISP-DM model.
• Summarization of existing CRISP-DM extensions.
• Development of the GCRISP-DS framework with dynamic phase interactions.

Main Results:

• The GCRISP-DS framework allows dynamic interactions between phases to address data and model issues.
• It emphasizes business understanding and data quality for achieving business objectives.
• The framework enhances model improvements and reusability while minimizing robustness issues.

Conclusions:

• The GCRISP-DS framework offers a customizable and flexible approach to data science.
• It effectively addresses practical challenges in implementing data-driven models.
• This enhanced framework supports the achievement of higher business objectives through robust data science practices.