Despite remarkable advances in artifificial intelligence (AI), many organizations continue to struggle with using these technologies to take advantage of their data and address business problems. Development of machine learning (ML) solutions in enterprise involves tackling a wide range of complexities with respect to requirements elicitation, design, development and deployment of such solutions. This process includes tackling challenges such as identifying the right business needs and use-cases, converting those needs into ML tasks and problems, specifying data requirements and transformation needs, selecting algorithms and assessing trade-offs, deploying and integrating ML models with business processes, and ensuring continuous alignment of ML applications with business strategies, amongst others. In spite of the necessity and relevance of conceptual modeling and requirements engineering approaches to the process, not much research has been done in this area.
GR4ML is a conceptual modeling framework for requirements elicitation, design, and development of machine learning solutions. The framework includes three modeling views, representing different aspects of a solution and viewpoints of different roles involved in the development of such systems:

Business View

This view provides a systematic way of revealing machine learning requirements by findings user personas, business strategies, decisions, along with business questions. The key question addressed by this view is: Who need the analytics system, for what and why?

Analytics Design View

It support designing the analytics solution by finding analytics types (predictive, descriptive, or prescriptive) and algorithms for the business questions at hand. The key question addressed by this view is: What algorithms are needed to generate the required insights?

Data Preparation View

It provides a systematic way of designing the data preparation workflows by representing database tables, attributes and relationships. The key question addressed by this view is: How would you need to prepare data to be consumed by algorithms?

The views an be constructed using a top-down, a bottom-up, or a hybrid approach. The three views are linked to each other to ensure the business and analytics alignment. Let's start by seeing a toy example here.