fbpx

Data Science Operationalization: What the heck is it?

By Walter Paliska

Data Science Operationalization Defined Data science operationalization, in concept, is simple enough: Take Machine Learning (ML) or Artificial Intelligence (AI) models and move them into production (or operational) environments. In the words of Gartner Sr. Analyst Peter Krensky, data science operationalization is the "...application and maintenance of predictive and prescriptive models..." In practice, however, operationalizing ML and AI models can be a complicated and often overwhelming challenge. In a broader concept, one of the biggest challenges of operationalization is that AI and ML models get integrated with systems that contain live data that changes quickly. For example, if your model is designed to predict customer churn, your data science operationalization process needs to be integrated with your CRM system to predict churn effectively as your data volumes grow. What makes data science operationalization so hard? There are four critical aspects of data science operationalization that make it challenging to implement.…

AutoML and Beyond – Part 1

By dotData

With AutoML trending in data science, our CEO spoke at #Ai4Finance on data preparation, aggregating tables, feature engineering, the #AutoML process, and AutoML’s missing gaps.  We’ll post the Conclusion / Part 2 next Thursday.  Video: Part 1 – AutoML and Beyond. Share On [social_warfare ] Related Articles

Are You Ready For Full-cycle AutoML on Python? – Part 2

By Sachin Andhare

conclusion from last week...Part 2 Beyond AutoML : Data Science Automation  While the rise of AutoML platforms has provided for faster execution of "test and learn" ML development, it has also brought about additional challenges. In most ML and data science projects, ML development is only one part of the process. The earlier stages of the process that require handling multiple raw tables and manipulating them based on in-depth domain knowledge to create flat, aggregated feature tables is a far more complicated and time-consuming challenge. The data and feature engineering process in enterprise data science has to deal with such different data as relational, transactional, temporal, geo-locational, and text data, which never starts from a single, flat, aggregated and cleansed table. Data science automation provides for a full-cycle automation process that includes data and feature engineering, in addition to standard AutoML. The ability to automatically generate features from massive and…