Understanding the Predictive Analytics Lifecycle

Conclusion

Success is better decision making. Previously with low volumes of data, intuitive decision making would work. As the data size has grown to incredible proportions, human ability to make completely intuitive decisions has been reduced. As a result, data-driven decision making has become more prevalent to ensure a reasonable path for success. This situation makes sense as it is easy to see that data are not diminishing but rather increasing.

These data-driven decisions are based often on quantitative models created using a typical closed-loop process: a cycle. The cycle described in this book includes:

• Problem definition and identification
• Design and build of an analytical framework, if there isn’t one available
• Data management, reporting, and visualization
• Analysis to produce models
• Execution and testing
• Feedback

Each iteration adds more knowledge to the model. Each step in the cycle is presented in a single chapter. Let us review key concepts of each chapter.

One of the problems organizations face when looking at the predictive analytics life cycle is that in many instances the professionals choosing the software and framework to conduct analytics do not understand the entire cycle and end up choosing disparate “tools” as opposed to holistic solutions. This creates issues for the researchers, such as delays accessing the diverse data, difficulties comparing discoveries, incompatibilities with previous results, challenges operationalizing the analytic models, and ...