Data feedback loops are critical for machine learning platforms according to research firm Wikibon. In his latest research Professional Alert, an elaboration on the role of digital business platforms in machine learning explored in his previous Alert, “Digital Business Platform for Machine Learning Apps”,  Wikibon Big Data & Analytics Analyst George Gilbert takes a close look at the data feedback loops that are central to machine learning. He finds that data wrangling and defining the variables or features that drive the model are critical to establishing successful analysis and should be entrusted to data scientists.

The data that feeds machine analysis often comes from sources that aren’t curated, and that means that preparing data requires human expertise. Defining the correct variables and data features with which to separate valuable data from noise is critical. Data that merely correlates with a model’s answers without contributing to the accuracy of those answers can result in less accurate or even misleading results.

Once these tasks are completed, however, the model can be largely automated, running as a virtuous loop that drives increasingly accurate analysis. This is a case were investment up front pays big dividends down the road in terms of better guidance for critical business decisions.

Wikibon Premium subscribers can read the research here.  To learn about subscribing, look here.

Image by werner22brigitte via Pixabay