Abstract
Data analytics is a key to unlock the untapped power of knowledge hidden in the trenches of big data in the age of information and intelligence. However, not all analytics efforts can achieve the desired outcomes and impacts. A successfully executed data analytics project relies on the application of an effective process model that facilitates multidisciplinary collaborations and balance technical efficiency with organizational effectiveness. This article introduces an integrated process model for data science and machine learning including its theoretical foundation and step-by-step descriptions and guidelines. The goal of this article is to provide concise, easy-to-follow guidance for data professionals and domain experts to apply this novel process model in their day-to-day collaborative and iterative analytics effort.
TopBackground
Data analytics is a key to unlock the untapped power of knowledge hidden in the trenches of big data in the age of information and intelligence. However, not all analytics efforts can achieve the desired outcomes and impacts. A successfully executed data analytics project relies on the application of an effective process model that facilitates multidisciplinary collaborations and balance technical efficiency with organizational effectiveness. Despite the abundance of existing analytics process models, many analytics professionals and project teams choose to use their own, or not to use any at all (Piatetsky-Shapiro, 2014). This low adoption rate of process models and the lack of a universal model inhibit the maturity and growth of the analytics profession in satisfying the increasing demand for data analytics.
Key Terms in this Chapter
Process Model: A standardized practice pattern or framework such as steps, associated tasks and activities, principles, and guidelines for planning, organizing, and executing a human endeavor. In modern society, human endeavors typically involve the use of technologies and the collaboration of people.
Design Science: A emerging research methodology focusing on the creation of artifacts (methods, techniques, and tools) and the evaluation of their performance in the physical, social, and cultural environment. Like a software application, a process model can be considered as an artifact and created and evaluated by following design science methodology.
Knowledge Discovery in Databases (KDD): A term coined by Piatetsky-Shapiro in 1989 and later was adopted as a data analytics process for discovering useful knowledge from a collection of data proposed by Fayyad, Piatetsky-Shapiro, and Smyth in 1996.
Decision Support Systems (DSS): A software system or application that supports managerial decision making, usually by modelling problems and employing quantitative models for solution analysis.
Sample-Explore-Modify-Model-Assess (SEMMA): A data analytics process model developed by the SAS Institute and is incorporated in its SAS Enterprise Miner software product. SEMMA represents six steps in data mining.
Business Intelligence (BI): A conceptual framework or a software application for managerial decision support. It combines technical architecture, databases, data warehouses data marts, analytical tools, data visualizations.